DRY GRANULATION VIA ROLLER COMPACTION : INVESTIGATION …hss.ulb.uni-bonn.de/2018/5074/5074.pdf · However, scaling up roller compaction processes is still not fully understood, identifying

DRY GRANULATION VIA ROLLER COMPACTION:

INVESTIGATION ON SCALE UP STRATEGIES

INTEGRATING PROCESS PARAMETERS AND

CRITICAL MATERIAL ATTRIBUTES

DISSERTATION

ZUR ERLANGUNG DES DOKTORGRADES (DR. RER. NAT.)

DER MATHEMATISCH-NATURWISSENSCHAFTLICHEN FAKULTÄT

DER RHEINISCHEN FRIEDRICH-WILHELMS-UNIVERSITÄT BONN

vorgelegt von

Robert Schmidtke

aus Düsseldorf

Dezember 2017

Angefertigt mit Genehmigung der Mathematisch-Naturwissenschaftlichen Fakultät der

Rheinischen Friedrich-Wilhelms-Universität Bonn

Prüfungskommission

Prof. Dr. K. G. Wagner (Erstgutachter)

Prof. Dr. A. Lamprecht (Zweitgutachter)

Prof. Dr. G. Bendas (Fachnahes Mitglied)

Prof. Dr. A. Schieber (Fachfremdes Mitglied)

Tag der Promotion: 04.05.2018

Erscheinungsjahr: 2018

Auszüge aus der Arbeit wurden an folgender Stelle vorab veröffentlicht:

Schmidtke R., Schröder D., Braun M., Wagner K.G.:

Dry granulation: Scale-up of roller compaction processes based on quality attributes of

ribbons and tablets

Poster, APV World Meeting 2016 Glasgow

Schmidtke R., Schröder D., Menth J., Staab A., Braun M., Wagner K.G:

Prediction of solid fraction from powder mixtures based on single component compression

analysis

International Journal of Pharmaceutics, Volume 523, Issue 1, 15 May 2017, Pages 366-375,

ISSN 0378-5173, http://doi.org/10.1016/j.ijpharm.2017.03.054

Schmidtke R., Schröder D., Braun M., Wagner K.G.:

Dry granulation: Scale - up of a roller compaction process - Adapted process settings to

achieve same product quality at different scales

Poster, APV World Meeting 2018 Granada

http://doi.org/10.1016/j.ijpharm.2017.03.054

Für meine Familie

Danksagung

Die vorliegende Arbeit entstand in der Zeit vom Februar 2014 bis zum Dezember 2017 in der

Abteilung Pharmazeutische Entwicklung der Boehringer Ingelheim Pharma GmbH & Co. KG

in Biberach. Für die besondere Möglichkeit eine Dissertation in der pharmazeutischen

Industrie anzufertigen und für die stetige Förderung danke ich Herrn Dr. Schreder und Herrn

Dr. Braun.

Meinem Doktorvater Herrn Prof. Dr. K. G. Wagner danke ich für das interessante Projekt,

seine stetige Unterstützung in allen Phasen der Doktorarbeit und die spannenden

Diskussionen, die mir immer wieder neue Blickwinkel eröffnet haben.

Weiterhin bedanke ich mich bei Herrn Prof. Dr. A. Lamprecht für die Anfertigung des

Zweitgutachtens, sowie bei Herrn Prof. Dr. G. Bendas und Herrn Prof. Dr. A. Schieber für ihr

Mitwirken in der Prüfungskommission.

Besonders bedanken möchte ich bei:

Dr. Daniela Schröder, meiner Betreuerin vor Ort, für unsere intensiven wöchentlichen

Diskussionen und die wertvollen Ratschläge, die dazu beigetragen haben den richtigen Weg

zu gehen.

Dr. Andrea Staab, Dr. Michael Braun für die Unterstützung, die perfekten

Arbeitsbedingungen und die Einbindung in die alltägliche Projektarbeit, die es mir ermöglicht

hat neue Erfahrungen zu sammeln.

Dr. Ragna Hoffmann für die interessanten Einblicke in ein Formulierungsentwicklungslabor

und die interessanten Diskussionen über neue Herangehensweisen in der

Formulierungsentwicklung.

Dr. Jens Borghardt für das mühevolle Korrekturlesen und die amüsanten Abende in Biberach.

Allen Kollegen der verschiedenen Labore der Prozess- und Formulierungsentwicklung, die

durch ihre Unterstützung, durch die unkomplizierte Zusammenarbeit zum Gelingen dieser

Arbeit beigetragen haben und immer an meine neuen Ideen geglaubt haben.

Vielen Dank an Florian und all meine Freunde, die mich während der letzten Jahre durch ihre

Gespräche, Diskussionen und diverser Wochenendtermine, abseits der pharmazeutischen

Technologie, unterstützt und auf andere Gedanken gebracht haben.

Meiner Familie danke ich für die stetige Förderung, die Unterstützung und den Glauben an

mich. Ohne euch wäre das nicht möglich gewesen.

Ganz besonders danke ich dir Patricia, für die Unterstützung in den letzten Jahren, die

unzähligen Gespräche, dein unglaubliches Verständnis für mich und deine grenzenlose Liebe.

Dadurch war diese Arbeit erst möglich.

I

TABLE OF CONTENTS

TABLE OF CONTENTS ......................................................................................................... I

ABBREVIATIONS AND SYMBOLS ................................................................................... V

1. INTRODUCTION ................................................................................................. 1

2. OBJECTIVES ........................................................................................................ 2

3. THEORETICAL ASPECTS AND MODEL DEVELOPMENT ...................... 5

3.1 ROLLER COMPACTION ................................................................... 5

3.2 POWDER COMPACTION AND SOLID FRACTION THEORY ..... 7

3.2.1 Powder compaction - Definitions .................................................... 7

3.2.2 Solid fraction of the ribbon as critical quality attribute for roller

compaction ....................................................................................... 8

3.2.3 Reduced tabletability caused by particle size enlargement effect,

lubricant, porosity and work hardening effect ............................... 10

3.3 PREDICTION OF SOLID FRACTION BASED ON

COMPRESSION ANALYSIS FOR TABLETS ................................ 13

3.3.1 Theoretical considerations Percolation, Kawakita and

exponential model - Mathematical model development ................ 14

4. RESULTS & DISCUSSION ............................................................................... 19

4.1 MATERIAL ATTRIBUTES AND METHOD COMPARISON

FOR SOLID FRACTION MEASUREMENTS OF RIBBONS ........ 19

4.1.1 Material attributes of raw materials and blends ............................. 19

4.1.2 Comparison of throughput, mercury porosimetry and

GeoPycnometry method to determine the solid fraction of

ribbons ........................................................................................... 26

4.2 COMPARISON OF TWO ROLLER COMPACTORS OF

DIFFERENT SCALE AT SAME PROCESS SETTINGS ................ 31

4.2.1 Formulation impact on the solid fraction within one scale ............ 31

4.2.2 Comparison of the solid fraction between different scales ............ 34

4.2.3 Particle size distribution of granules .............................................. 35

4.2.4 Influence of granules on tablet attributes ....................................... 45

4.2.5 Summary ........................................................................................ 57

II

4.3 ADAPTED PROCESS SETTINGS OF DIFFERENT SCALES TO

ACHIEVE SIMILAR PRODUCT QUALITY ................................... 59

4.3.1 Achieving the same solid fraction of ribbon by using adapted

process parameter settings ............................................................. 59

4.3.2 Particle size distribution and porosity of granules ......................... 62

4.3.3 Tabletability and compressibility influenced by material

attributes of granules and ribbons .................................................. 69

4.3.4 Summary ........................................................................................ 73

4.4 INVESTIGATION OF SOLID FRACTION DISTRIBUTION

ALONG THE ROLL WIDTH BETWEEN DIFFERENT SCALES

VIA NIR AT-LINE ............................................................................ 75

4.4.1 Method evaluation to determine solid fraction along the roll

width by GeoPycnometer and NIR ................................................ 76

4.4.2 Development of an at-line NIR method for solid fraction

measurements of ribbons ............................................................... 88

4.4.3 Scale up approach – Comparison of solid fractions of ribbons of

a Metformin formulation at two scales .......................................... 94

4.4.4 Solid fraction distribution along the roll width between different

scales by NIR ................................................................................. 96

4.4.5 Summary ........................................................................................ 99

4.5 MODEL DEVELOPMENT – PREDICTING SOLID FRACTION

OF A TABLET ................................................................................. 101

4.5.1 Application of models – Excipients ............................................. 102

4.5.2 Prediction of solid fraction – Mixtures ........................................ 108

4.5.3 Summary ...................................................................................... 110

5. MATERIALS ..................................................................................................... 111

5.1 MATERIALS ................................................................................... 111

5.2 FORMULATIONS ........................................................................... 113

5.2.1 Placebo composition - Chapter 4.1, 4.2 and 4.3 .......................... 113

5.2.2 API composition - Chapter 4.4 .................................................... 113

5.2.3 Placebo composition- Chapter 4.5 ............................................... 114

6. MANUFACTURING & ANALYTICAL METHODS ................................... 115

6.1 MANUFACTURING AND TECHNOLOGIES .............................. 115

III

6.1.1 PREPARATION OF MIXTURES AND PROCESS FLOW

CHARTS ...................................................................................... 117

6.2 ANALYTICAL METHODS ............................................................ 127

6.2.1 Raw material and granules ........................................................... 128

6.2.2 Ribbon – Solid fraction ................................................................ 132

6.2.3 Tablet ........................................................................................... 139

7. SUMMARY ........................................................................................................ 141

8. APPENDIX ........................................................................................................ 145

8.1 ANALYTIC DATA .......................................................................... 145

8.1.1 CHAPTER 4.1 MATERIAL ATTRIBUTES AND METHOD

COMPARISON FOR SOLID FRACTION MEASUREMENTS

OF RIBBONS .............................................................................. 145

8.1.2 CHAPTER 4.2 COMPARISON OF TWO ROLLER

COMPACTORS OF DIFFERENT SCALE AT SAME

PROCESS .................................................................................... 146

8.1.3 CHAPTER 4.4 INVESTIGATION OF SOLID FRACTION

DISTRIBUTION ALONG THE ROLL WIDTH BETWEEN

DIFFERENT SCALES VIA NIR AT-LINE ............................... 149

8.1.4 MODEL DEVELOPMENT – PREDICTING SOLID

FRACTION OF A TABLET ....................................................... 150

8.2 STATISTICAL TESTS .................................................................... 151


COMPARISON FOR SOLID FRACTION MEASUREMENTS

OF RIBBONS .............................................................................. 151

8.2.2 CHAPTER 4.2 COMPARISON OF TWO ROLLER

COMPACTORS OF DIFFERENT SCALE AT SAME

PROCESS .................................................................................... 152

8.2.3 CHAPTER 4.3 ADAPTED PROCESS SETTINGS OF

DIFFERENT SCALES TO ACHIEVE SIMILAR PRODUCT

QUALITY .................................................................................... 158

8.2.4 CHAPTER 4.4 INVESTIGATION OF SOLID FRACTION

DISTRIBUTION ALONG THE ROLL WIDTH BETWEEN

DIFFERENT SCALES VIA NIR ................................................ 161

IV

8.3 LIST OF TABLES ........................................................................... 162

REFERENCES ..................................................................................................................... 164

V

ABBREVIATIONS AND SYMBOLS

Abbreviation Definition

API Active pharmaceutical ingredient

CARB Sodium carboxymethylcellulose (AcDiSol®)

CPP Critical process parameters

DL Drug load

GeoPyc GeoPycnometer

LAC −Lactosemonohydrate (Tablettose 80®)

MacroPactor M1075-GMP-Polygran

MCC Microcrystalline cellulose (Avicel® PH102)

MET Metformin hydrochloride

MGST Magnesiumstearate (LIGAMED®)

NIR Near infrared reflectance

PCA Principal component analysis

PLS Partial least square regression

PSD Particle size distribution

RSD Relative standard deviation %

SD Standard deviation of mean

SF Solid fraction

TS Tensile strength

USP United States Pharmacopeial

Symbol Definition

Porosity

𝑥𝐸𝑥𝑐𝑖𝑝𝑖𝑒𝑛𝑡 Weight fraction of excipient in mixture [% w/w]

𝜎𝑡 Tensile strength [N/mm²]

𝜎𝑡𝑚𝑎𝑥 Maximal tensile strength [N/mm²]

π Pi

A Heckel - Intercept

a Kawakita constant

b Kawakita constant

C Degree of volume reduction

cr Curvature radius

D Diamter [cm]

d50 Median particle dimension [µm]

d,f,g Exponential constants

drolls Diamter rolls [cm]

VI

F Force [N]

h Height [cm]

hc Heigth calotte [cm]

k Heckel slope

m Mass [g]

𝑝𝑐 Critical concentration or percolation treshold (percolation theory)

P Pressure [MPa]

Py Heckel Yield pressure

q Critical exponent or compressibility exponent (percolation theory)

r Radius

R2 Correlation coefficient

S Proportional constant (percolation theory)

SFmax Maximal solid fraction

t Production time [min]

TSgranule Tensile strength tablets of granules [N/mm²]

TSpowder Tensile strength tablets of powder [N/mm²]

V0 Initial volume [cm³]

Vmin Minimal achievable volume [cm³]

VP Volume of tablet at applied pressure [cm³]

vrolls Velocity rolls [rpm]

W Central cylinder thickness [cm]

wrolls Width rolls [cm]

xp Compression pressure [MPa]

y Surface tension

θ Contact angle [°]

ρAPP Apparent density [g/cm³]

ρTRUE True density [g/cm³]

𝑋 System property (percolation theory)

𝑝 Site occupation or bond probability (percolation theory)

π Pi

y Surface tension liquid [dyn/cm]

θ Angle [°]

INTRODUCTION

1

1. INTRODUCTION

Granulation processes for solid oral dosage forms have been widely used as an intermediate

process step in the pharmaceutical industry. The advantages of a granulation process are

improved granule attributes such as flowability, compressibility, tabletability and less

segregation of the active pharmaceutical ingredient (API) and excipients, which contributes to

a better accuracy of metering, content uniformity and followed by a higher final product

quality [1]. The purpose of this technique is to agglomerate excipients with the API and to

obtain a desired product quality of e.g. tablets or capsules. Whenever this method is applied,

the quality of the final product is controlled by resulting attributes of granules, which are

influenced by the process parameters of the granulation technique. Wet and dry granulation

techniques represent the main techniques. During wet granulation, using binder solution to

build liquid bridges based on capillary and viscous force between particles, causes

agglomeration. In contrast, dry granulation is characterised by using mechanical force to

facilitate interparticulate bond formation. In comparison to wet granulation, dry granulation

provides the following advantages:

• suitable for water- or heat-sensitive APIs,

• simple to operate due to integrated process control mechanism [2],

• minimal energy consumption,

• increased bulk density of the product,

• feasible for implementation in a continuous manufacturing process [1,3].

Most commonly applied dry granulation method is roller compaction. The granulation unit is

equipped with two counter-rotating rolls, whereby a screw system (auger) conveys powder

into a compaction zone between these rolls. By applying a specific compaction force, a

compact is formed. An integrated mill afterwards grinds the formed ribbon, to obtain granules

for downstream processing, e.g. tableting. The main disadvantage of roller compaction is the

loss of tensile strength of tablets that are produced based on the granules in comparison to

tablets based on unprocessed material, caused by previously consumed plasticity (work

hardening phenomena), particle size enlargement of the granules and lubrication effects by

added lubricant [4–10]. Furthermore, in the pharmaceutical industry the change from small

development batches to commercial batches requires a transfer from a small scale to a larger

scale (scale up) to satisfy market demands. However, scaling up roller compaction processes

is still not fully understood, identifying the factors that finally determine reproducible tablet

quality remains to be accomplished. Investigating these factors, comprising material attributes

OBJECTIVES

2

and the impact of the scale on the final drug product quality will be shown in the course of

this thesis.

2. OBJECTIVES

The objective of this thesis was to investigate the effect of different roller compactor scales

over the whole process chain on the quality attributes of intermediate- and final products, i.e.

tablets, in order to develop a reliable scale up strategy.

At first two commonly used formulations for a roller compaction process with different

properties, containing different fractions of Microcrystalline cellulose (MCC) and pre-

agglomerated −Lactosemonohydrate (LAC), will be characterised to understand their impact

on the quality attributes for downstream processing (i.e. tableting). Furthermore, a comparison

of analytical techniques for the critical quality attribute solid fraction of a ribbon was

performed to find an appropriate method for this thesis. Based on these experiments, both

formulations were compacted, using equal process parameters at both scales to identify and

separate differences caused by material attributes (formulation) and scale. This was expected

to result in a profound new view on the scalability of a roller compaction process. Afterwards

a scale approach was proposed balancing the difference between various scales to achieve the

same quality attributes of a tablet, whereby the controversially discussed topic of the impact

of the solid fraction of ribbons, particle size distribution and porosity of granules on tablet’s

quality attributes was assessed. In this aspect, the influence of the solid fraction of the ribbon

was investigated in a more detailed view between scales using a newly developed near

infrared spectroscopy method. Moreover, solid fraction was an important impact factor, which

reinforced the development of theoretical models to predict the solid fraction for powder

mixtures based on single component compression analysis.

OBJECTIVES

3

Table 1 Overview objectives

Chapter Objective Summary

4.1 Characterisation of material attributes and method evaluation for the measurement of

the solid fraction of ribbons

4.2

Investigation of the impact of formulation attributes and different scales on

intermediate products ribbon, granules and tablets at different scales at same process

settings

4.2.5

4.3 Scale Model approach to balance observed differences between scales at adapted

process settings 4.3.4

4.4 Determination of the solid fraction distribution of ribbons along the roll width

between two scales via near infrared reflectance 4.4.5

4.5 Prediction of the solid fraction of tablets from powder mixtures based on single

component compression analysis 4.5.3

This thesis is structured in five chapters. The first three chapters (chapter: 4.1, 4.2, 4.3)

contain the impact of material attributes of two formulations and correlated differences caused

by the scale of the roller compactor. Three main quality attributes will be investigated: (1)

solid fraction of ribbons, (2) attributes of granules and (3) tablets. Finally, the solid fraction

between scale will be determined more detailed (4.4), followed by comparing three theoretical

models (4.5) to predict the solid fraction of mixtures for tablets.

OBJECTIVES

4

ROLLER COMPACTION

5

3. THEORETICAL ASPECTS AND MODEL DEVELOPMENT

3.1 ROLLER COMPACTION

Three main process areas can be distinguished for a roller compactor (see Figure 1):

1. Conveying system -> Powder

2. Compaction zone -> Ribbon

3. Milling system -> Granules

A powder blend is conveyed by an agitator, feed auger and tamp auger (force-feed system)

between counter-rotating rolls into the compaction zone. Depending on the manufacturer,

various orientations of the conveying system are used [3].

Figure 1 Schematic drawing – Roller compactor

The compaction zone is divided into a slip, nip region and release region [1]. Initially, in the

slip region particle rearrangement and de-aeration occurs, followed by an increasing specific

compaction force in the course along the nip region, where the velocity of the particles

becomes equal to the rotating rolls [11,12]. Different surfaces of the rolls (e.g. knurled,

smooth) improve the friction between rolls and powder, which ease the dragging of powder

into the nip region. Powder densification occurs in this region. Particle deformation (plastic

deformation) and particle fragmentation takes place, whereby new bonds occur between

particles, caused by more contact points and newly formed surfaces. Maximum pressure is

achieved on the powder right before the minimal gap between the two rolls [1]. A seal system

(cheek plates or rim rolls) encloses the compaction zone to prevent side seal leakage of the

ROLLER COMPACTION

6

powder. A ribbon is being released from the compaction zone to the milling system. Here, the

released ribbon is sheared and sliced against a sieve mesh by an oscillating granulator to

obtain granule particles.

Two roller compactors were used for this thesis: A MiniPactor® and a M1075-GMP-

Polygran®. Both machines are products of the company Gerteis (Gerteis Machinen +

Processengineering AG, Switzerland). The roller compactors are characterised by an inclined

conveying system with a feed auger and tamp auger (see Figure 1). These machines provide

an automatic gap control system, whereby the feed of the augers is automatically adjusted if

the gap exceeds the defined value, which results in a constant powder supply and thus to a

low fluctuation of the gap. The ratio of the feed/tamp auger can be set as process parameter.

Samples of the ribbons are taken after achieving steady state conditions for the gap. Both

machines were equipped with cheek plates as side seal system. Details of the used process

parameter for each experiment are provided in 6.1.1. Figure 2 gives an overview of all design

aspects and process parameters for a roller compaction process.

Figure 2 Design aspects and process parameters of a roller compactor

All construction aspects between both machines are equal, except for the roll width. The

MiniPactor® (small scale) has a roll width of 25 mm, in contrast to the M1075-GMP-

Polygran® (large scale), which has a roll width of 50 mm and thus increases the throughput.

M1075-GMP-Polygran® will be referred to as MacroPactor® in this thesis.

POWDER COMPACTION AND SOLID FRACTION THEORY

7

3.2 POWDER COMPACTION AND SOLID FRACTION THEORY

3.2.1 Powder compaction - Definitions

Single solid dosage forms are mostly prepared by compressing powder to tablets. Applying

pressure on powder causes a volume reduction, whereby bondings are formed by plastic

deformation, particle fragmentation, resulting in new available surfaces for bondings, and a

formed compact is obtained (e.g. ribbon, tablet). The compact can be characterised measuring

their tensile strength (TS) and solid fraction (SF).

Solid fraction (SF) is the possible volume reduction related to its true density (lowest possible

volume), and represents how dense a compact is compressed.

𝑆𝑜𝑙𝑖𝑑 𝑓𝑟𝑎𝑐𝑡𝑖𝑜𝑛 =𝜌𝐴𝑝𝑝𝑎𝑟𝑒𝑛𝑡 𝑑𝑒𝑛𝑠𝑖𝑡𝑦

𝜌𝑇𝑟𝑢𝑒 𝑑𝑒𝑛𝑠𝑖𝑡𝑦=

𝑚

𝑉𝑃𝑚

𝑉𝑚𝑖𝑛

Eq. (1)

m = mass of powder/tablet [g]; Vp = volume at applied pressure [cm³]; Vmin = minimal volume [cm³]

Solid fraction is a quality attribute of ribbons manufactured by a roller compaction (see 3.2.2)

and tablets. The solid fraction of a tablet is a critical quality attribute (CQA) as it correlates to

disintegration and dissolution [13–15] (i.e. faster dissolution and disintegration with lower

solid fraction). The reason is that a low SF combined with a high porosity facilitates liquid

penetration into tablets (Porosity () = 1- solid fraction).

The second attribute of a compact is the tensile strength (TS), which is determined by

measuring the required radial force to break the compact, whereby the geometry of the

compact is considered for calculation (generalised breaking force) [16]. TS as parameter of

mechanical strength of tablets has a major relevance for tablets during downstream processing

due to mechanical impact during coating, transportation and packaging [17].


8

Figure 3 Definitions – Tabletability, Compressibility and Compactibility

Measurements of the tensile strength and solid fraction allow characterising materials and

granules with regard to their tabletability, compressibility and compactibility (see Figure 3).

Compressibility plots enable a view on the consolidation process of the material under

pressure. Mathematical equations can be applied for compressibility plots to get a deeper

insight into the physical behaviour of the densification process (e.g. Heckel, Kawakita, see

6.2.1.5). Tabletability shows the mechanical strength dependent on pressure, and

compactibility is a combination of both values as the tensile strength is depicted on a certain

degree of densification.

3.2.2 Solid fraction of the ribbon as critical quality attribute for roller compaction

In respect to the regulatory guidance of the ICH Q8 (R2) Pharmaceutical development [18]

application of Quality-by-Design (QbD) approach means to understand the process in depth,

which enables to “built in” quality into the product by a process design space. Based on this,

intermediate critical material attributes (iCMA), critical process parameters (CPP) and critical

quality attributes (CQA) are examined during the development of a drug product. These three

aspects define the control strategy of a drug product. For roller compaction the solid fraction

of ribbon is a key intermediate critical quality attribute [5,8], as it shows an impact on particle

size distribution (PSD) [5,19–24], porosity of granules [25] and the tensile strength of tablets

[5–7,9,26]. Different process parameters like the speed of the conveying system, roll speed,

gap width and specific compaction force have an impact on these quality attributes [20].


9

As an example, a short dwell time of the ribbon is caused by a low speed of the rolls or

conveying system and results to a low solid fraction [21,27–29]. In contrast, a high solid

fraction is obtained by an increased specific compaction force or a lower gap width [5]. A

control strategy for a roller compaction process contains these process parameters, but the

resulting solid fraction of a ribbon reflects a combination of all process parameters together.

Hence, it is not surprising that various authors have investigated the solid fraction in depth.

Studies reported determining the solid fraction comprise different analytical techniques:

• X-ray µCT [30],

• ultra sonics [31],

• geometrical method [32],

• modified geometrical method [27,33],

• throughput [5,26,27,34]

• GeoPycnometer [8,17,35],

• mercury porosimetry [22],

• light transmission [36],

• oil absorption [22,37],

• throughput method [5,26,27,34],

• buoyancy method [33],

• near infrared reflectance [32,38],

• near infrared reflectance – chemical imaging [21,22]

It was demonstrated that a non-uniform solid fraction distribution of the ribbon along the roll

width occurs [21,30,31,36]. Two design aspects of the roller compactor are identified to cause

this effect: Conveying and side seal system. A periodical sinusoidal solid fraction distribution

is obtained due to the feeding pressure of the last flight of the tamp auger [39]. Different side

seal systems are used: cheek plates and rim rolls. Rim rolls led to a higher solid fraction at the

edges [27], whereby a higher solid fraction at the centre was obtained using cheek plates [40],

which can be diminished by internal powder lubrication [41].

In contrast, only a few authors measured and investigated the effect of different scales on the

solid fraction. Unfortunately, the few reports on scale were controversial. Alleso et al. (2016)

[37] stated that the scale has no impact on the solid fraction of the ribbon. In contrast to that,

Shi et al. (2016) [42] and Ana Pérez Gago et al. (2017) [43] recognized a scale dependent

influence. Unfortunately, neither the solid fraction distribution nor a potentially resulting

impact on the granule and tablet quality were discussed or published.


10

3.2.3 Reduced tabletability caused by particle size enlargement effect, lubricant,

porosity and work hardening effect

Potential reasons are described in literature, explaining the main disadvantage of roller

compaction process, namely the loss of tensile strength (tabletability), comprise four main

reasons:

• work hardening phenomena,

• particle size enlargement effect,

• porosity of granules,

• added amount of Magnesium stearate to the granules,

The term of work-hardening was first introduced by Malkowska et al. (1983) [44] who

compressed excipients, milled and re-tableted the obtained granules. They observed a loss of

tensile strength compared to the unprocessed excipients. It was stated that work-hardening

means “resistance to deformation” of material, which can be described as a partial loss of

their ability to build a new network of bonds. The probability of building bonds increases with

a higher contact area between particles, which is dependent on the specific surface of particles

and thus to the particle size distribution (PSD) [45]. As a coarser granule size occurs after

roller compaction, which has a reduced specific surface, various authors tried to distinguish if

the work-hardening or the particle size enlargement would affect the loss of tensile strength

after roller compaction.

Sun et al. (2006) [46] tableted a small and a coarse lubricated sieve fraction of

Microcrystalline cellulose (MCC) after multiple compaction cycles and observed within one

sieve fraction a decrease of the specific surface area, whereby the coarser sieve fraction

provided always a lower tensile strength. They concluded that the particle size enlargement

effect caused the loss of tensile strength. In contrast, Herting et al. (2008) [47] used

unlubricated sieve cuts of MCC and stated that the loss of tensile strength is impacted due to

an effect of both work-hardening and particle size enlargement. Wu et al. (2007) [48]

determined that the particle size enlargement effect can be considered as negligible by using

brittle components, as these components consolidate under an extensive fracturing during

tableting into smaller particles with a higher specific surface. Compared to brittle

components, He et al. (2007) [49] showed that MCC (plastic consolidation) is more sensitive

towards addition of Magnesium stearate (MGST) regarding loss of tensile strength. Mosig et

al. (2015) [10] tried to investigate these effects: particle size enlargement, work hardening and

added lubricant for a plastic (MCC) and brittle ( − Lactosemonohydrate = LAC)

components. They confirmed the study of He et al. (2007) [49] as they detected that the loss


11

of tensile strength after roller compaction of MCC was enhanced due to lubrication.

Additionally, an effect of work-hardening and particle size enlargement was verified.

Furthermore, no loss of tensile strength for LAC could be observed either by lubrication,

particle size enlargement or after roller compaction, due to the brittle fracturing behaviour,

which results into smaller particles with new unlubricated surfaces. Nordstrom et al. (2015)

[25] proposed an interesting new aspect. Highly porous granules disintegrate into their

primary particles during tableting, which diminishes the impact of the granule size on the loss

of tensile strength. Thus, the granule porosity is an important factor, which has an impact on

granules’ attributes. Recently, Sun et al. (2016) [50] published a mini review on this topic,

where they concluded that all factors have to be considered for a roller compaction process.


12

PREDICTION OF SOLID FRACTION BASED ON COMPRESSION ANALYSIS FOR

TABLETS

13

3.3 PREDICTION OF SOLID FRACTION BASED ON COMPRESSION

ANALYSIS FOR TABLETS

The solid fraction is commonly understood as an important aspect of formulation design as it

directly influences tensile strength, disintegration of tablets, dissolution time of tablets, drug

product stability [13,14,51,52] and serves as scale up characteristic [53]. Thereby, it can be

considered as critical quality attribute (CQA) [15].

Hence, predicting of the solid fraction based on single compression analysis of commonly

used excipients would be highly beneficial for development purposes. The prediction may

serve as a systematic guidance for the formulator to select appropriate excipients depending

on the active pharmaceutical ingredient to build quality into the drug product according to the

Quality by Design approach.

Compression analysis of pharmaceutical powders has been reported by various authors [54–

60]. The most frequently used compression models are Heckel [61] and Kawakita [62], which

provide a physical interpretation of the volume reduction process of powders dependent on

the applied pressure. The Heckel model for pharmaceutical powders is derived from

compression experiments of metal powders and assumes that the consolidation (plastic

deformation) follows first-order kinetics, which results in the Heckel equation Eq. (2), where

is the porosity of the compact and k the reciprocate of the Yield pressure.

− ln( ) = 𝑘 ∗ 𝑃 + 𝐴 Eq. (2)

= porosity; k = slope Heckel; 𝑃 = compression pressure; A = intercept

A linear course of the Heckel plot at increasing pressure indicates plastic deformation.

However, in contrast to metal powders, pharmaceutical powders display additionally to plastic

deformability, particle rearrangement and/or elastic deformation, which leads to deviations

from the linear course of plastic deformation behaviour (see 6.2.1.5.1). In this context the

Heckel plot shows a curvature in the lower pressure region [54]. Duberg et al. (1986) decided

to divide the Heckel plot into 3 phases for pharmaceutical powders to reach a better

applicability of the Heckel equation: particle rearrangement or fracturing, elastic or plastic

deformation and decompression. Consequently, no single Heckel equation will be able to

describe the compressibility of pharmaceutical powders appropriately for the entire range of

the applied compression pressure. As for implementing, a prediction model for solid fraction

for the widest possible range of compression pressure requires a model, which parameters

describe the compression behaviour of the respective pressure range, the Heckel equation was

excluded for this study. In contrast to the Heckel model, the Kawakita Eq. (3) equation


TABLETS

14

assumes that particles under compression pressure (P) are in an equilibrium and the product of

the pressure term and volume term is constant [57].

𝑃

𝐶=

𝑃

𝑎+

1

𝑎𝑏 Eq. (3)

C = degree of volume reduction; P = compression pressure [MPa]; a = Kawakita constant; b = Kawakita

constant

Consequently, a linear course is obtained when plotting P/C vs. compression pressure, where

C is the degree of volume reduction. The Kawakita parameters a and b-1 are determined by

linear regression and likely to deliver appropriate results over the whole compression

pressure. Promising results were found using the Kawakita equation to predict the

compressibility of a tablet successfully [58,64,65]. Another potential model is the percolation

model. Usually employed to elucidate the governing property of a material in a powder

mixture dependent on its volume fraction, it can also be applied to describe a property,

dependent on the fracture or extent of a process parameter. In this case, the tablet property of

interest would be the solid fraction while the process parameter would be the compression

pressure. A successful implementation of this concept has been demonstrated by various

authors [66–70]. A sudden property change of a tablet will only be observed if the particle

rearrangement is completed and an infinite cluster can be formed [66]. This sudden change

(percolation threshold) is not considered by the simplification of the compressibility by the

Kawakita equation. Therefore, percolation theory was applied for predicting the solid fraction

as a function of the compression pressure. This adapted model has the potential to improve

predictions for the solid fraction of a ternary mixture compared to Kawakita.

3.3.1 Theoretical considerations Percolation, Kawakita and exponential model -

Mathematical model development

Two theoretical models Percolation and Kawakita will be evaluated for model application. An

exponential model is added to elucidate whether the two-parametrised models with theoretical

background are superior in terms of predictability of solid fraction compared to a model

without parametrised variables.

3.3.1.1 Percolation

In the course of the percolation theory Eq. (4), tableting of powder is considered as forming

site- and bond clusters in a lattice [66]. After particle rearrangement an infinite cluster is

formed throughout the whole tablet, and the particles cannot disintegrate into their primary

particles again [66]. Before this percolation threshold is reached, the voids of the powder bed


TABLETS

15

are filled by particle rearrangement. The percolation threshold is typically between the tapped

density and the bulk density, where a first compact is formed and the interparticulate bonding

starts, so that isolated clusters (finite clusters) are combined to form an infinite cluster

throughout the tablet, which corresponds to a massive property change in the system.

Basic formula of percolation phenomena:

𝑋 = 𝑆 ∗ (𝑝 − 𝑝𝑐)𝑞 [66] Eq. (4)

Percolation formula for tensile strength:

𝜎𝑡

𝜎𝑡𝑚𝑎𝑥= 𝑆 ∗ (𝑝 − 𝑝𝑐)𝑞 [69] Eq. (5)

X = system property; S = proportional constant or scaling factor; 𝑝 = site occupation or bond probability;

𝑝𝑐

= critical concentration or percolation threshold; 𝑞 = critical exponent; 𝜎𝑡 = tensile strength [N/mm²];

𝜎𝑡𝑚𝑎𝑥= maximal tensile strength [N/mm²]

In theory, the basic power law is only valid near the percolation threshold in a lattice [66].

Kuentz & Leuenberger found that it is possible to use the percolation theory for a broader

range regarding modified Young’s modulus and tensile strength [68,69,71] . Different authors

evaluated the applicability of the percolation theory for tensile strength by considering the

system property X as tensile strength, 𝑝 as relative density (or solid fraction), and 𝑞 as

fractural exponent of the tablet [66–69]. They defined the percolation threshold (𝑝𝑐) as a

minimum of solid fraction which leads to a mechanical strength or as a “critical volume

fraction in a continuum percolation” [69]. The value found for the percolation threshold was

between the relative bulk density and the tapped density for the excipients [69]. The

theoretical value of the critical exponent 𝑞 can be calculated by applying the Bethe lattice

approximation or mean field theory. For mechanical strength it is defined as constant with a

value of 2.7 [67]. Some authors found that 𝑞 can differ from theoretical values for tensile

strength in a binary system [67–69]. Related to this knowledge, 𝑞 was defined as variable

parameter to predict the solid fraction referring to it as compressibility exponent (q)

throughout the manuscript. The solid fraction (SF) is normalised by the highest detected value

for SF (𝑆𝐹𝑚𝑎𝑥), which leads to:

𝑆𝐹

𝑆𝐹𝑚𝑎𝑥= 𝑆 ∗ (𝑥𝑝 − 𝑝𝑐)

𝑞 Eq. (6)

SF = solid fraction; 𝑆𝐹𝑚𝑎𝑥 = measured maximal solid fraction; S = scaling factor; xp = compression pressure

[MPa]; 𝑝𝑐= Percolation threshold; 𝑞 = compressibility exponent

Eq. (6) was used for fitting the model variables S, 𝑝𝑐 and 𝑞 using the measured SF and 𝑆𝐹𝑚𝑎𝑥

values.


TABLETS

16

3.3.1.2 Kawakita

The Kawakita equation [72] Eq. (7) is based on the volume reduction of powder (𝐶) under

compression pressure. 𝐶 is defined as the degree of volume reduction (engineering strain) at

applied pressure 𝑃. 𝑉0 is the initial in-die volume of the powder, and 𝑉𝑃 is the volume of the

powder at applied pressure.

𝐶 = 𝑉0−𝑉𝑃

𝑉0=

𝑎∗𝑏∗𝑃

1+𝑏∗𝑃 Eq. (7)

𝑎 and 𝑏 are Kawakita compression parameters which can be determined by linear regression

using the linearized form of Eq. (8) [73,74],

𝑃

𝐶=

𝑃

𝑎+

1

𝑎𝑏 Eq. (8)

𝑎 represents the maximal strain or degree of compression at maximal pressure (𝐶𝑚𝑎𝑥). The

inverted 𝑏-1 (1/𝑏) describes the pressure to reach 𝑎/2, which can be correlated to the plasticity

(Yield pressure, Heckel) and initial compressibility of single ductile granules [74] or can be

seen more simplified as deformation capacity [75]. The parameters a and b were determined

by applying linear regression on a plot of P/SF vs. P. Kawakita’s 𝑎 is equal to 1/slope and 𝑏-1

is equal to slope multiplied with y-intercept. Considering the determination of 𝐶 and therefore

the initial volume (𝑉0) can have an important influence on the results [76,77]. There are three

methods described for the determination of 𝑉0. The first is to measure 𝑉0 between 1-2 MPa

[77] or at the lowest measurable pressure. The second is to define 𝑉0 based on bulk density

which delivers better results compared to the first method [78]. The third is to determine

initial density using nonlinear regression with three parameters [77,79]. These estimate are

highly dependent on the method of determination (process and user) and hence, prone to

errors. Thus, the resulting Kawakita parameters are difficult to compare between various

research laboratories.

Therefore, a modified approach to determine 𝑉0 was chosen, where 𝑉0 should only be

material-dependent and particle size dependence is negligible. Similarly, to the Heckel

approach, the reference volume was the lowest possible volume, i.e. the volume at maximum

density or true density. Subsequently, 𝐶 was related to solid fraction and 𝑉0 changed to Vmin,

which was measured by Helium-pycnometry, an easy and reliable determination method. The

values of Vmin and VP are defined in Eq. (9) and Eq. (10) :


TABLETS

17

𝜌𝑇𝑅𝑈𝐸 =𝑚

𝑉𝑚𝑖𝑛; 𝑉𝑚𝑖𝑛 =

𝑚

𝜌𝑇𝑅𝑈𝐸 Eq. (9)

𝜌𝐴𝑃𝑃 =𝑚

𝑉𝑃 ; 𝑉𝑃 =

𝑚

𝜌𝐴𝑃𝑃 Eq. (10)

𝜌𝑇𝑅𝑈𝐸 = true density [g/cm³]; 𝜌𝐴𝑃𝑃 = apparent density [g/cm³]; m = mass of tablet [g]

where 𝑉𝑃 represents the volume of the powder at applied pressure, and 𝜌𝐴𝑃𝑃 is the

corresponding apparent density in Eq. (10). As 𝑉𝑚𝑖𝑛 is always smaller than 𝑉𝑃, 𝐶 is derived

as:

𝐶 = 𝑉𝑚𝑖𝑛

𝑉𝑃 Eq. (11)

If Eq. (9) is combined with Eq. (10):

𝑉𝑚𝑖𝑛

𝑉𝑃=

𝑚

𝜌𝑇𝑅𝑈𝐸𝑚

𝜌𝐴𝑃𝑃

or 𝑉𝑚𝑖𝑛

𝑉𝑃=

𝜌𝐴𝑃𝑃

𝜌𝑇𝑅𝑈𝐸 Eq. (12)

Considering Eq. (12) and Eq. (1) leads to:

𝐶 = 𝑆𝐹 =𝜌𝐴𝑃𝑃

𝜌𝑇𝑅𝑈𝐸=

𝑎∗𝑏∗𝑃

1+𝑏∗𝑃 Eq. (13)

Where SF is the solid fraction.

Thus, the modified Kawakita parameter a can be considered as the maximum solid fraction at

an examined compression pressure range that is achievable for an excipient by tableting.

Therefore, the following modified Kawakita formula is proposed and used:

𝑃

𝑆𝐹=

𝑃

𝑎+

1

𝑎𝑏 Eq. (14)

3.3.1.3 Exponential

In addition to the prediction of solid fraction by the modified Kawakita model and the

Percolation model, a simple exponential function Eq. (15) was used to predict the solid

fraction without a mechanistic model behind it, i.e. the variable d, f and g are adapted, non-

parametrised variables.

𝑆𝐹 = 𝑑 + 𝑓 ∗ 𝑒(𝑔∗𝑃) Eq. (15)

The variables 𝑑, 𝑓 and 𝑔 were fitted by linear regression using a compressibility plot (solid

fraction vs. compression pressure [MPa]).

3.3.1.4 Additive rule

Ramaswamy et al. (1970) demonstrated that the volume of a mixture follows an additive rule

of the volume of single components, under the condition that the single components have the

same particle size [80].


TABLETS

18

𝑆𝑜𝑙𝑖𝑑 𝑓𝑟𝑎𝑐𝑡𝑖𝑜𝑛𝑀𝑖𝑥𝑡𝑢𝑟𝑒 = 𝑥𝑀𝐶𝐶 ∗ (𝑀𝑜𝑑𝑒𝑙) + 𝑥𝐿𝐴𝐶 ∗ (𝑀𝑜𝑑𝑒𝑙) + 𝑥𝐶𝐴𝑅𝐵 ∗ (𝑀𝑜𝑑𝑒𝑙) Eq. (16)

𝑥𝐸𝑥𝑐𝑖𝑝𝑖𝑒𝑛𝑡 = weight fraction (% w/w) of excipient in mixture

This hypothesis was applied by various authors to predict the porosity, percolation threshold,

or tensile strength of mixtures [75,79,81–83] and was shown to be adequate for excipients

with different particle size. This approach was applied for all three models to predict the solid

fraction, where 𝑥𝐸𝑥𝑐𝑖𝑝𝑖𝑒𝑛𝑡 is the weight fraction (% w/w) of the respective single excipient of

the mixture. Magnesium stearate was not included in the calculations as it was used as

lubricant at a constant level for both, the single excipients and the mixtures. It was necessary

to determine the true densities of the materials to apply the calculation of the solid fraction

according to Eq. (1) and to calculate the true densities of the mixtures Eq. (17).

For the mixtures the true density was calculated by the weight fraction of MCC, LAC and

Sodium carboxymethylcellulose(CARB) divided by 99.5, as described by Gupta et al. (2005)

[84], to reduce the number of input parameters for the models.

ρ𝑇𝑅𝑈𝐸𝑀𝑖𝑥𝑡𝑢𝑟𝑒

=𝑥𝑀𝐶𝐶

99.5∗ (𝜌𝑇𝑅𝑈𝐸 𝑀𝐶𝐶) +

𝑥𝐿𝐴𝐶

99.5∗ (𝜌𝑇𝑅𝑈𝐸 𝐿𝐴𝐶) +

𝑥𝐶𝐴𝑅𝐵

99.5∗ (𝜌𝑇𝑅𝑈𝐸 𝐶𝐴𝑅𝐵) Eq. (17)

𝜌𝑇𝑅𝑈𝐸= true density [g/cm³]; 𝑥𝐸𝑥𝑐𝑖𝑝𝑖𝑒𝑛𝑡 = weight fraction (% w/w) of excipient in mixture

MATERIAL ATTRIBUTES AND METHOD COMPARISON FOR SOLID FRACTION

MEASUREMENTS OF RIBBONS

19

4. RESULTS & DISCUSSION

4.1 MATERIAL ATTRIBUTES AND METHOD COMPARISON FOR SOLID

FRACTION MEASUREMENTS OF RIBBONS

4.1.1 Material attributes of raw materials and blends

In order to understand the results of a roller compaction process, it is necessary to characterise

the excipients and blends with respect to their attributes (see 4.1.1.1) and compression

behaviour (compressibility, tabletability and compactibility, see 3.2.1).

4.1.1.1 Raw material properties

Table 2 Material attributes - Raw materials and blends

Composition

True

density

[g/cm³]

Bulk

density

[g/ml³]

Tapped

density

[g/ml]

Hausner

ratio

(2500 taps)

Particle size

distribution d50

Microcrystalline cellulose 1.5565 0.21 0.27 1.32 87

−Lactose-monohydrate 1.5417 0.63 0.79 1.27 157

Sodium carboxymethylcellulose 1.5934 0.50 0.71 1.38 61

MCC 2:1 LAC 1.5472 0.48 0.58 1.23 103

MCC 1:1 LAC 1.5447 0.52 0.63 1.23 110

Metformin hydrochloride 1.3559 - - - -

MET 21 1.5066 0.50 0.62 1.23 98

MCC = Microcrystalline cellulose; LAC = − Lactosemonohydrate; MET 21 = Metformin with drug load 21 %

For all excipients and blends (MCC 2:1 LAC, MCC 1:1 LAC) their true density, bulk/tapped

density, particle size distribution (d50) were characterised. In Table 2, an overview of the

material attributes is provided.



20

4.1.1.2 Compression analysis

In literature, there are a several methods described for the analysis of the powder densification

process to define the properties of the material [60–63,85–88]. Compressibility, tabletability

and compactibility plots were considered in this thesis (see 3.2.1). All single components

(LAC, MCC, CARB) and blends (MCC 2:1 LAC, MCC 1:1 LAC) were tableted by a single

punch tablet press (FlexiTab). Despite the knowledge that MGST can influence the

compression behaviour [89], 0.5 % MGST was added to the excipients to guarantee same

process conditions compared to the lubricated blends processed by a roller compactor.

4.1.1.2.1 Compressibility

An incremental displacement transducer system traced the displacement of the punches during

compression in order to obtain force displacement data in the range of 0 – 235 MPa.

Figure 4 Compressibility – “In- die” measurement excipients and blends,

CARB = Sodium carboxymethylcellulose,

LAC = − Lactosemonohydrate, MCC = Microcrystalline cellulose –

Image section: Compression pressure 20 – 70 MPa, Exceedance of

LAC’s solid fraction by MCC’s solid fraction at 66 MPa



21

Consequently, the solid fraction of the tablet was plotted against the compression pressure

(see Figure 4). LAC achieved a higher solid fraction at a low compression pressure compared

to MCC. LAC consisted of spherically pre-agglomerated particles and had a higher bulk

density of 0.63 g/cm³ (working/starting density) compared to the density of 0.21 g/cm³ of

MCC. The compression behaviour between LAC and MCC is different. LAC has a brittle

compression behaviour [90,91]. In contrast, MCC consolidates by plastic deformation under

pressure [60], explaining the steeper slope of the compressibility plot and exceeded LAC’s

solid fraction at around 0.77 ( 66 MPa). CARB also undergoes plastic consolidation, but

compared to MCC to a lower extent. The short decrease of compression pressure at 22 MPa

(1.7 kN) was attributed to the switching threshold from pneumatic to hydraulic pressure of the

FlexiTab. Considering the compressibility of both blends, MCC 1:1 LAC had a higher solid

fraction at lower compression pressure compared to MCC 2:1 LAC, which is attributed to the

high fraction of LAC. At around 0.74 SF ( 66 MPa) MCC 2:1 LAC exceeded the solid

fraction of MCC 1:1 LAC, indicating a higher impact of the plastic consolidation of MCC.

The course of the solid fraction plot versus compression pressure of all mixtures was in

agreement with the sum of the properties of the single components in relation to their fraction

within the mixture. This finding is consistent with results published by various authors

[58,80,92,93] who have shown that the volume reduction of a blend (eq. solid fraction)

follows an additive rule of the volume reduction of single excipients in a blend (see 3.3.1.4).

4.1.1.2.1.1 Compressibility equation – Heckel

Yield pressure (Heckel) determination was done in a pressure range between 20 MPa –

120 MPa for unprocessed material as this range corresponds to a solid fraction of the blends

between 0.60 – 0.80, which represents an expected range for a solid fraction of a ribbon at dry

granulation [17]. A low Yield pressure illustrates a good plastic consolidation, which is

defined as a low resistance against material deformation.

LAC showed the highest Yield pressure of 188.68 MPa compared to all other investigated

excipients and blends. This was attributed to the brittle deformation behaviour of LAC [90,91].

In contrast, MCC showed the lowest Yield pressure of 86.21 MPa, which was caused by the

good plastic consolidation. The Yield pressure of the blends (107.53 MPa MCC 2:1 LAC,

119.05 MPa MCC 1:1 LAC) increased with a higher fraction of the brittle LAC [56,93].

Determined values are consistent with literature [54,56], knowing the limitation that Heckel

plots are difficult to compare between various research laboratories [54]. All correlation

coefficients were above 0.98. Assuming that particle rearrangement is not fully completed at



22

20 MPa and that normal operating range for a Heckel plot is 50 MPa [63,93,94], the fitting

result were adequate.

Figure 5 Heckel plot – “In-die” measurement excipients and blends, CARB =

Sodium carboxymethylcellulose , LAC = − Lactosemonohydrate,

MCC = Microcrystalline cellulose - Image section: Pressure range for

Yield pressure determination (linear regression)

Table 3 Results Heckel – Excipients and blends

MCC LAC CARB MCC 2:1 LAC MCC 1:1 LAC

Slope 0.0116 0.0053 0.0079 0.0093 0.0084

Py (1/slope) 86.21 188.68 126.58 107.53 119.05

R² 0.9863 0.9855 0.9988 0.9877 0.9867

Py = Yield pressure Heckel; R2 = coefficient of correlation; MCC = Microcrystalline cellulose;

LAC = − Lactosemonohydrate; CARB = Sodium carboxymethylcellulose



23

4.1.1.2.2 Tabletability

In Figure 6 the course of the tensile strength depending on compression pressure for each

excipient and blend is shown. The tensile strength increased with rising compression pressure.

At low compression pressure, determination of the tensile strength of LAC was not possible

using an automatic tablet tester because the resulting tablets were too fragile.

Figure 6 Tabletability – Excipients and blends, mean (n = 6), error bars

(standard deviation of mean), CARB = Sodium

carboxymethylcellulose, LAC = − Lactosemonohydrate,

MCC = Microcrystalline cellulose

The maximal of tensile strength decreased in the order MCC > CARB > LAC, which was a

result of the compression behaviour of good plastic consolidation (MCC, CARB) and brittle

fracturing (LAC) during compression. The plastic flow of MCC and CARB resulted in strong

bonds and consequently led to a higher tensile strength. LAC showed a nearly linear increase

of the tensile strength [95], however at overall a low tensile strength level. Considering the



24

blends, tensile strength decreased with a higher fraction of LAC. These differences were

persistent over the whole range of compression pressure.

4.1.1.2.3 Compactibility

A good compactibility represents a high tensile strength at a low degree of solid fraction.

MCC achieved the highest tensile strength at a comparable solid fraction ( 0.90) followed by

MCC 2:1 LAC > MCC 1:1 LAC > LAC. CARB achieved only a maximal solid fraction of

about 0.83, which however was sufficient to achieve a good tensile strength (see Figure 7).

Figure 7 Compactibility – Excipients and blends, mean (n = 6), error bars

(standard deviation of mean), CARB = Sodium

carboxymethylcellulose, LAC = − Lactosemonohydrate,

MCC = Microcrystalline cellulose

Considering the course of the blends, it is obvious that MCC 2:1 LAC always showed a

higher tensile strength at a similar solid fraction compared to MCC 1:1 LAC. In other words

MCC 2:1 LAC needed less pressure (eq. solid fraction) to reach a higher tensile strength [96]

because of the higher proportion of MCC and its good plastic consolidation under pressure.



25

4.1.1.2.4 Summary

In summary, a high plastic deformation under pressure resulted in a good tabletability and

high compactibility. MCC showed the highest values over the whole range of investigated

compression pressures. A higher fraction of LAC in a mixture resulted in a lower tabletability

and compactibility. Considering the compressibility plot, LAC reached a higher solid fraction

than MCC at a low compression pressure. Heckel Yield pressure results confirmed the course

of the compressibility plots.

Table 4 Summary - Compression analysis excipients and blends – Descending order of

maximal achievable tabletability and compactibility

Composition Py Tabletability Compactibility

Microcrystalline cellulose 86.21

Sodium carboxymethylcellulose 126.58

MCC 2:1 LAC 107.53

MCC 1:1 LAC 119.05

−Lactose-monohydrate 188.68

Py = Yield pressure Heckel; MCC = Microcrystalline cellulose; LAC = − Lactosemonohydrate



26

4.1.2 Comparison of throughput, mercury porosimetry and GeoPycnometry

method to determine the solid fraction of ribbons

As a first step, it is important to identify an appropriate method to analyse the ribbons.

Different methods have been described by authors for the determination of solid fraction for a

ribbon (see 3.2.2). A few techniques were compared. These methods can be distinguished by

the determination of the “in-gap” solid fraction within the gap during compaction, (three

throughput methods) and “out of gap” solid fraction, after the ribbon is released (throughput +

height measurement, GeoPycnometry, mercury porosimetry) (see 6.2.2). All samples were

produced within one roller compactor (MacroPactor) at 2, 4, 6 and 8 kN/cm (MCC 2:1 LAC).

In gap methods - Three “in-gap” methods [5,26,27] were compared. As depicted in Figure 8,

solid fraction increased up to 6 kN/cm. Comparing the “in-gap” methods of Herting et al.

(2007) [26], Gamble et al. (2010) [5] and Peter (2010) [27] differences could not be noticed

(see Figure 8).

Figure 8 “In-gap” solid fraction ribbon – Mean (n = 5), error bars (standard

deviation of mean) – Comparison of three approaches to calculate the

solid fraction of ribbons by weighing the granule throughput



27

The volume calculation of the ribbon was based on the equation of Herting et al. (2007) [26].

An extension of this equation was done by Gamble et al. (2010) [5] who took the voids of the

knurled surface of the rolls also into account. Peter (2010) [27] did a correction by

multiplying the circumference by half of the gap. Considering these three equations (see

6.2.2.1) the calculated volume was fixed and defined by the geometry of roller compactor and

process settings (gap width). Thus, only the produced mass of granules per minute (g/min)

could have an impact on the results. Small relative standard deviations for the solid fraction

between 0.72 – 2.44 % were obtained. Higher specific compaction forces result in higher solid

fraction and higher tensile strength of the ribbons [17,23]. As a result, the ribbons require

more time to get milled by the granulator [97,98]. Throughput per minute dropped, and

therefore the calculated solid fraction was reduced, too (Figure 8, see equations 6.2.2.1). This

effect was most pronounced for the specific compaction force of 8 kN/cm.



28

Out of gap methods - The “out of gap” porosity was measured by the proposed approach of

Nkansah et al. (2008) [34], by GeoPycnometry and mercury porosimetry. Three methods were

compared. In order to consider the elastic relaxation of the ribbons, measurements were done

48h following compaction. For Nkansah et al. (2008) [34] mean height of the ribbons was

measured five times by a micrometer screw. In the range of 2-4 kN/cm, differences could not

be identified between these methods. At 6 kN/cm the solid fraction of Nkansah et al. (2008)

[34] was lower than the others. The difference increased to 0.08 for ribbons at 8 kN/cm (see

Figure 9).

Figure 9 “Out of gap” solid fraction ribbon – Mean, error bars (standard

deviation of mean), methods: GeoPycnometer (n = 5), Mercury

porosimetry (n = 3), Nkansah et al. (2008) [34] (n = 5)

A higher residence time of the ribbons in the granulator caused a decrease in solid fraction for

the obtained ribbons. The observed maximal relative standard deviation were smaller for

mercury porosimetry (0.78 – 1.75 %) and for the GeoPycnometry (0.29 – 0.42 %) compared

to Nkansah’s approach (0.72 – 2.44 %). The differences between the GeoPycnometry and

mercury porosimetry were in a range between -0.61 and 1.21 %. A significant difference



29

(p 0.05) between both methods was not observed (see APPENDIX 8.2.1 T-Test). Hence,

there is neither a systematic nor significant deviation between GeoPycnometry and mercury

porosimetry. In summary the “in-gap” solid fraction did not include the elastic recovery as

only the gap setting of the roller compactor is a variable for the calculation (see 6.2.2.1).

Therefore, higher values were delivered for the solid fraction compared to the “out of gap”

solid fraction (see Figure 10). The throughput methods (“in-gap” & “out of gap”) delivered

reasonable results at lower specific compaction force, but they were prone to errors at higher

specific compaction forces.

Figure 10 MacroPactor – Comparison of analytical methods for solid fraction of

ribbon, mean, error bars (standard deviation of mean)

Due to the low maximal relative standard deviation and robustness of the measurement, the

mercury porosimetry and GeoPycnometry are the analytical methods of choice. Considering

sample preparation, only small pieces (2* 25 mm * 10 mm) can be analysed by mercury

porosimetry, because of the small volume of sample chamber of the Dilatometer compared to

the GeoPycnometry (6 * 25 mm * 30 mm), which will give a more representative result.

Analysis time is around 1 h for mercury porosimetry compared to 15 - 20 min of the



30

GeoPycnometry. Hence, for both reasons the GeoPycnometry was the analytical method for

determination of ribbon’s solid fraction in this thesis.

COMPARISON OF TWO ROLLER COMPACTORS OF DIFFERENT SCALE AT SAME

PROCESS SETTINGS

31

4.2 COMPARISON OF TWO ROLLER COMPACTORS OF DIFFERENT

SCALE AT SAME PROCESS SETTINGS

Two formulations MCC 2:1 LAC and MCC 1:1 LAC were compacted at four specific

compaction levels (2, 4, 6, 8 kN/cm) at different scales, while process parameters were kept

identical (see Table 13). The main difference between these roller compactors was the roll

width (25 mm MiniPactor, 50 mm MacroPactor). Solid fraction of the ribbons, attributes of

granules and tablets were examined in particular regarding following two aspects:

1. Impact of the formulation (MCC 2:1 LAC, MCC 1:1 LAC)

2. Influence of different scales (MacroPactor, MiniPactor)

4.2.1 Formulation impact on the solid fraction within one scale

MacroPactor - Solid fraction increased with increasing specific compaction force [99]. The

solid fraction range increased from 0.62 to 0.82 for applied force of 2 kN/cm to 8 kN/cm.

Figure 11 Solid fraction ribbon – MacroPactor, mean (n = 5), error bars

(standard deviation of mean)


PROCESS SETTINGS

32

This solid fraction was in agreement with previously published normal operating range for a

roller compaction process [17]. The relative standard deviation of means was small (0.30 % –

1.05 %, see Figure 11). A T-test indicated significant difference (p 0.05) between these two

formulations (see APPENDIX 8.2.2). A higher fraction of LAC led to a higher solid fraction

of the ribbon, which was consistent with findings of various authors [4,23].

Compression behaviour of the spherically shaped pre-agglomerated LAC particles indicated a

higher solid fraction. LAC had higher bulk and tapped density (0.63 g/ml³, 0.79 g/ml³)

compared to MCC (0.21 g/ml³, 0.27 g/ml³). Additionally, LAC needed less pressure for

particle rearrangement because of the brittle compression behaviour [90,91], which is

characterised by the fracturing of pre-agglomerated primary particles under pressure. A higher

fraction of LAC causes a consolidation to a higher solid fraction [4]. This behaviour was

confirmed by “In-die” tableting measurements of single components previously (see 4.1.1.2,

Figure 4). The compaction process of a tablet press and a roller compactor is not precisely

comparable because the roller compactor has a longer dwell time according to Hilden et al.

(2011) [100]. But a higher fraction of LAC caused the formulation of MCC 1:1 LAC to reach

a higher solid fraction compared to MCC 2:1 LAC up to a compression pressure of 66 MPa

for “In-die” tableting (see 4.1.1.2, Figure 4). Exceeding that pressure MCC 2:1 LAC will

result in higher solid fractions. Initially, the solid fraction of LAC was higher at low

compression pressure, which was reflected by a high bulk and tapped density. Considering the

course of the profiles in Figure 11, the differences decreased from 0.022 to 0.005 between

both formulations from low to high specific compaction force. A higher fraction of MCC

induces a stronger increase of the solid fraction by plastic consolidation [101], which took

place at higher specific compaction forces and compensated the initial lower solid fraction of

formulation MCC 2:1 LAC. A comparable behaviour was observed for unprocessed blends by

“In-die” tableting (see 4.1.1.2, Figure 4).


PROCESS SETTINGS

33

MiniPactor – MiniPactor showed similar results for the ribbons compared to the

MacroPactor for both formulations. An equal impact of the formulation attributes on the

resulting solid fraction of ribbons was observed (Figure 12). T-test indicated significant

difference (p 0.05) (see APPENDIX 8.2.2). Relative standard deviation 0.59 % - 3.90 %

was higher compared to the MacroPactor. A explanation could be a lower steady state powder

supply into the gap for the small scale, which led to a higher relative standard deviation.

However, the relative standard deviation was still in an acceptable range. Difference between

formulations decreased from 0.039 to 0.008 (2 kN/cm to 8 kN/cm).

Figure 12 Solid fraction ribbon – MiniPactor, mean (n = 5), error bars (standard

deviation of mean)


PROCESS SETTINGS

34

4.2.2 Comparison of the solid fraction between different scales

Comparing the solid fraction of ribbons at two different scales for each formulation, a

difference between both scales was observed, despite identical process parameters were

chosen (see Figure 13).

Figure 13 Comparison solid fraction ribbon - MacroPactor/MiniPactor –

MCC 2:1 LAC/MCC 1:1 LAC, mean (n = 5), error bars (standard

deviation of mean)

The deviation between the two scales was between 18.18 % to 9.27 %, the lowest difference

occurred at highest specific compaction force of 8 kN/cm. A T-Test confirmed a significant

difference (p 0.05) of the solid fraction between both scales (see APPENDIX 8.2.2). In

literature, only limited information is available about direct comparisons of two different

scales with the same formulation at same process settings. Alleso et al. (2016) [37] used only

MCC and obtained no difference for four of five batches. They concluded that the solid

fraction of ribbons at different scales is equal. Shi et al. (2016) [42] recognized a higher solid

fraction for the ribbons that were produced at a larger scale, which is in accordance to Figure


PROCESS SETTINGS

35

13. Recently, Ana Pérez Gago et al. (2017) [43] published that dependent on the used

formulation, a different scale can impact the solid fraction. A clear explanation for this

difference was however, not provided. Thus, a contradictory picture exists in literature and no

author has shown the effect of scale on granules and likewise on resulting tablets. In

conclusion, a combined analysis of the granules (4.2.3) and tablets (4.2.4) is still missing to

better understand the influence of the observed difference at different scales.

4.2.3 Particle size distribution of granules

4.2.3.1 Impact of formulation attributes on particle size distribution within one scale

MacroPactor - It is common understanding that an increased specific compaction force (eq.

solid fraction) results in coarser granules [19–22]. Results of the d50 of the granules can be

seen in Figure 14.

Figure 14 d50 granules – MacroPactor – MCC 2:1 LAC/MCC 1:1 LAC, mean

(n = 3), error bars (standard deviation of mean)


PROCESS SETTINGS

36

The d50 of MCC 2:1 LAC increased from 83 µm to 133 µm and the bulk density from

0.51 g/cm³ to 0.56 g/cm³ for specific compaction force of 2 kN/cm to 8 kN/cm, which

reflected a densification of the formulation.

Considering Figure 15, the fraction of fine particles ( 63 µm) increased up to 6 kN/cm,

compared to the unprocessed blend (0 kN/cm). Fracturing of pre-agglomerated particles of

LAC under pressure causes smaller particles with a higher surface in respect to unprocessed

LAC [19,23,102]. The higher surface can also be considered as an increased bonding

capacity.

Figure 15 PSD granules – MacroPactor - MCC 2:1 LAC, triangle (mean, n = 3),

error bars (standard deviation of mean), colour indicates different

specific compaction forces [kN/cm]

The effect of disintegration into fine particles by pre-agglomerated particles and brittle

fracturing was reduced at a higher solid fraction of the ribbon caused by an increased impact

of the plastically consolidated MCC. LAC particles can fill the voids between fibrous MCC

particles [4] to build a ribbon to resist the shear stress of the milling process of the granulator

(see Figure 15).


PROCESS SETTINGS

37

The d50 of the particle size distribution (PSD) of the formulation MCC 1:1 LAC increased

from 79 µm to 210 µm and the bulk density increased from 0.53 g/cm³ to 0.59 g/cm³. At a

same specific compaction force, it would be expected, that the higher fraction of the MCC in

formulation MCC 2:1 LAC would result in a stronger ribbon characterised by a higher tensile

strength compared to MCC 1:1 LAC (see 4.1.1.2.2). Hence, a higher d50 occurs [23], but this

was not be observed for 8 kN/cm (see Figure 14).

Figure 16 Side seal leakage of unprocessed material above cheek plates –

MacroPactor – MCC 2:1 LAC at 8 kN/cm

A leakage at the side seal system was observed for MCC 2:1 LAC at 8 kN/cm. Unprocessed,

smaller material slipped above the compaction zone (see Figure 16). With a higher amount of

smaller particles the d50 of MCC 2:1 LAC was reduced. This explained the lower d50 of

133 µm for MCC 2:1 LAC compared to 210 µm for formulation MCC 1:1 LAC at 8 kN/cm.

Furthermore, no increase of the d50 could be observed between 6 kN/cm and 8 kN/cm (see

Figure 14). Hence, MiniPactor results will be more representative for a reliable formulation

comparison for PSD of granules within one scale as no side seal leakage was observed.


PROCESS SETTINGS

38

MiniPactor – An increased d50 of the granules was obtained at rising specific compaction

force. MCC 2:1 LAC showed a strong increase of the d50 from 88 µm to 136 µm compared to

MCC 1:1 LAC (70 µm to 105 µm, see Figure 17). Same process parameters resulted in a

higher d50 for MCC 2:1 LAC. This was caused by the higher amount of MCC, which is

correlated with a higher plastic deformation, resulting in a higher tensile strength (see Figure

6), to resist the destructive load of the milling process of the granulator. LAC needed a higher

pressure (e.g. degree of consolidation) to gain the same level of tensile strength (see Figure 6).

Therefore, a higher d50 was obtained for MCC 2:1 LAC at the same specific compaction force

compared to MCC 1:1 LAC.

Figure 17 d50 granules – MiniPactor – MCC 2:1 LAC/MCC 1:1 LAC, mean (n =

3), error bars (standard deviation of mean)

Brittle behaviour and destruction of the fine pre-agglomerated LAC particles showed a lower

d50 at low specific compaction force compared to unprocessed blends. The d50 of the

unprocessed blends have been firstly exceeded after a specific compaction force of about

6 kN/cm (see Figure 17). Bulk density increased from 0.47 g/cm³ to 0.56 g/cm³ (2 kN/cm –


PROCESS SETTINGS

39

8 kN/cm, MCC 2:1 LAC) and from 0.50 g/cm³ to 0.60 g/cm³ (2 kN/cm – 8 kN/cm, MCC 1:1

LAC) with rising solid fraction, which indicated a densification of both formulations.

Comparing both formulations, a higher fraction of fine particles were found for MCC 1:1

LAC compared to MCC 2:1 LAC (see Figure 18, 63 µm). Although the formulation MCC

1:1 LAC reached a higher solid fraction of the ribbon within one scale (see Figure 13).

Figure 18 PSD granules– MiniPactor at 2, 4, 6 and 8 kN/cm –

MCC 2:1 LAC/MCC 1:1 LAC, triangle (mean, n = 3), error bars


The formulation MCC 2:1 LAC had a higher ribbon compactibility (see 4.1.1.2.2), resulting

in a higher tensile strength of the ribbon [23], which enhanced the resistance against the shear

stress of the granulator, and resulted in a lower fraction of fine particles.

A higher solid fraction of the ribbon (MCC 1:1 LAC, see Figure 12) does not lead to a coarser

particle size if two different formulations are compared at identical process parameters within

one scale. Especially, if one contains a higher fraction of a brittle component (e.g. LAC). The

bonding strength between particles is essential for the PSD as it enhances the resistance

against the granulator and reduces the spall of small particles during granulation. MCC 1:1


PROCESS SETTINGS

40

LAC showed lower bonding strength (e.g. compactibility, see Figure 7), resulting in a higher

amount of finer particles, compared to MCC 2:1 LAC.

4.2.3.2 Comparison of particle size distribution between different scales dependent on

same formulation

MCC 2:1 LAC - PSD of granules at different scales showed only small differences. At low

specific compaction forces (2 kN/cm, 4 kN/cm) a similar d50 was found, compared to 6 kN/cm

or 8 kN/cm (see Figure 14, Figure 17), whereby 8 kN/cm can be excluded for evaluation,

because a side seal leakage of fine particles was observed for the MacroPactor (see Figure

16). A detailed view on the PSD showed only a small difference between the small and the

large scale (see Figure 19). A higher fraction of coarse granules was found at 6 kN/cm for the

MacroPactor. A higher solid fraction of the ribbon at the MacroPactor resulted in a similar

PSD of the granules at low solid fraction.

Figure 19 PSD granules - MacroPactor/MiniPactor at 2, 4, 6 and 8 kN/cm -

MCC 2:1 LAC, mean (n = 3), error bars (standard deviation of mean)


PROCESS SETTINGS

41

MCC 1:1 LAC - Different results were found for the formulation MCC 1:1 LAC. Differences

of the PSD at different scales were more distinctive at high specific compaction forces. The

d50 was comparable (see Figure 14, Figure 17) at low specific compaction forces (2 kN/cm, 4

kN/cm). A higher d50 was achieved for the MacroPactor at 6 kN/cm and 8 kN/cm compared to

the MiniPactor.

Figure 20 PSD granules – MacroPactor/MiniPactor at 2, 4, 6 and 8 kN/cm -

MCC 1:1 LAC, triangle (mean, n = 3), error bars (standard deviation

of mean)

A difference was noticeable for the fraction of fine particles between both scales (see Figure

20). The fraction of fine particles was smaller for the granules of the MacroPactor. This effect

was more pronounced for formulation MCC 1:1 LAC compared to MCC 2:1 LAC (see Figure

19, Figure 20), because a lower fraction of MCC reduced the compactibility of the

formulation to resist the shear stress of the granulator. This reduced compactibility was

enhanced due to the increased amount of brittle LAC (see 4.1.1.2.3). The effect of a higher

amount of LAC in the formulation, resulting in finer granules within one scale was previously

mentioned (see 4.2.3.1). However, the impact of a higher solid fraction of the ribbon at a


PROCESS SETTINGS

42

larger scale (MacroPactor, see Figure 13) on the PSD was even more pronounced for a

formulation with a lower compactibility (MCC 1:1 LAC). Thus, the higher solid fraction of

the MacroPactor caused coarser granules compared to the MiniPactor.

4.2.3.3 Impact of the milling process at different scales

A comparison of the milling step at both scales was done to prove if the solid fraction of the

ribbon would be the only impacting factor for the previously observed differences of PSD.

Blends of both formulations were tableted at different target solid fractions (MCC 1:1 LAC

0.64, 0.79; MCC 2:1 LAC 0.62, 0.77; see Table 25) to compare the milling process between

scales.

Figure 21 Comparison milling process scale – Granulator MacroPactor

One batch with the same solid fraction contained 2000 tablets, equally divided between the

two roller compactors to be milled in the granulator. Eight granules were analysed. During

tableting (FlexiTab®) the die wall was lubricated automatically by a press chamber coating

system [103] after every ten tablets to reduce the friction between powder and die wall and to

achieve a homogenous density distribution of the tablets [104]. The geometry of 20 tablets of

each batch was determined by an automatic tablet tester to calculate the observed solid

fraction (see APPENDIX 8.1.2).


PROCESS SETTINGS

43

Table 5 d50 milled tablets – MacroPactor/MiniPactor – MCC 2:1 LAC/MCC 1:1 LAC

Formulation Target

solid fraction

MacroPactor

d50 [µm] SD

MiniPactor

d50 [µm] SD

MCC 2:1 LAC

0.62 99 5 100 2

0.77 163 7 164 10

MCC 1:1 LAC

0.64 86 0 88 2

0.79 139 7 146 3

n = 3; d50 = Median particle dimension; SD = Standard deviation of mean

No significant differences of the d50 (p 0.05) between both scales were observed (see Table

5, see T-Test APPENDIX 8.2.2). A detailed examination of the PSD plot confirmed the

impression. The milling process between the two scales had no significant impact on the

particle size distribution of granules (see Figure 22).

Figure 22 PSD milled tablets of 0.62, 0.64, 0.77 and 0.79 solid fraction -

MiniPactor/MacroPactor – MCC 2:1 LAC/MCC 1:1 LAC, mean



PROCESS SETTINGS

44

Thus, it can be concluded that milling a product (e.g. tablets) with the same solid fraction at

different scale resulted in an equal PSD. Consequently, the previously demonstrated different

PSD of the granules was the result of the diverse solid fraction of the ribbon at different scales

(see Figure 19, Figure 20). This is in agreement with literature; authors have identified the

solid fraction of the ribbon as key quality attribute for downstream processing

[4,5,8,17,26,99]. Now, for the first time, differences of the PSD between two different scales

applying identical process parameters were demonstrated.

In summary, it can be concluded that a high solid fraction (eq. specific compaction force)

provided coarser granules (see Figure 14, Figure 17) in connection with a low amount of fine

particles. A larger scale (MacroPactor) achieved a higher solid fraction compared to the

smaller scale (MiniPactor) for both formulations at same process settings (Figure 13).

Differences of the PSD of the granules between scales were more distinctive at high values

( 0.70) of the solid fractions. This was in particular seen for the formulation with a high

amount of the brittle LAC, because formulation MCC 1:1 LAC had a lower compactibility

than MCC 2:1 LAC (see 4.1.1.2.3). That enhanced the effect of a lower solid fraction of the

ribbons on PSD for the smaller scale. In terms of the PSD of the resulting granules, MCC 1:1

LAC showed a higher susceptibility to different solid fraction caused by various roller

compactors.


PROCESS SETTINGS

45

4.2.4 Influence of granules on tablet attributes

All granules and unprocessed blends were tableted with equal process parameters. In Figure

23 (MacroPactor) and Figure 24 (MiniPactor) the tensile strength of tablets vs. compression

pressure of each scales and for both formulations is plotted.

4.2.4.1 Tabletability of formulations within one scale

Figure 23 Tabletability – MacroPactor - MCC 2:1 LAC/MCC 1:1 LAC, mean

(n = 50), error bars (standard deviation of mean), colour indicates

different specific compaction force [kN/cm]

MacroPactor & MiniPactor - One way ANOVA with post hoc Bonferroni analyses was

performed to evaluate the impact of the specific compaction force on the tensile strength. A

significant difference (p 0.05) was identified for all test samples.

(except 14 of 140 showed no difference; for details see APPENDIX 8.2.2). At an increasing

specific compaction force a decrease of the tensile strength could be found for all

formulations independent of scale. The observed loss of tabletability with increasing specific


PROCESS SETTINGS

46

compaction force after dry granulation is in agreement with various authors [4–7,9]. The loss

of tabletability is attributed to various reasons like work hardening [7,9,44,105,106], granule

size enlargement [46,47], an enhanced effect resulting from adding MGST to the granules

[48,49] and a different granule porosity [25]. Work hardening means that plastic deformable

excipients, which were loaded by pressure before, partially lost their ability to undergo plastic

deformation to build a network of bonds. In addition, building a coherent network like a tablet

is influenced by the ability to undergo bonds, which is correlated to the surface area of

individual particles [45]. This surface is affected by particle size and particle surface covered

by MGST on the particle. Brittle and porous excipients/granules tend to fracture during the

tableting process into smaller particles, creating new surfaces. These surfaces are not covered

by MGST and have larger specific surfaces (smaller particles) to build new contact points

[56], resulting in stronger bonds compared to unfractured coarse granules, which are covered

by a MGST layer.

Figure 24 Tabletability – MiniPactor - MCC 2:1 LAC/MCC 1:1 LAC, mean (n =

50), error bars (standard deviation of mean), colour indicates different

specific compaction forces [kN/cm]


PROCESS SETTINGS

47

MCC 2:1 LAC reached a higher tensile strength in comparison to MCC 1:1 LAC at both

scales (see Figure 23, Figure 24) because of the higher fraction of MCC, which showed a

better tabletability and compactibility compared to LAC (see 4.1.1.2).

The loss of tabletability with increasing specific compaction force is more pronounced for

MCC 2:1 LAC as the ability of MCC to undergo plastic deformation again, decreases with

rising specific compaction force (work hardening) [44,46]. This effect enhances by internal

lubrication, whereby the surfaces are covered with a layer of MGST [49]. Additionally, a

filming or coating of the MCC particles by MGST can be observed, whereby the whole

surfaces are covered by MGST, if high concentration of MGST and long blending duration

were chosen. Filming or coating of the whole particles can be excluded for these results as

only a total amount 1 % MGST and a low blending time was used (see Table 15, Table 21). In

contrast, LAC shows insensitivity towards MGST [10] and only a negligible loss of tensile

strength (low work hardening) compared to unprocessed LAC (0 kN/cm) after dry

granulation, which is caused by the ability of fracturing into smaller particles with a higher

specific surface [90,107].

Various authors [4,10,108,109] have shown a nearly linear increase of the tensile strength for

LAC after roller compaction over a range of compression pressure. Only a minor influence of

the specific compaction force of a roller compactor on the loss of tensile strength was found,

which was independent of particle size enlargement (lower surface area). In other words, a

high fraction of LAC (brittle) results in a higher reworkability of the formulation. MCC 1:1

LAC showed a linear course of the tensile strength at increased specific compaction forces

compared to low specific compaction forces (see Figure 23, Figure 24), which can be

correlated to a higher impact of the linear consolidation of the brittle LAC after dry

granulation, because less plastic consolidation of the MCC was available (work hardening).


PROCESS SETTINGS

48

A good illustration of the reworkability of a formulation is the reworkability index, firstly

introduced by Herting et al. (2007) [26]. He compared the achieved tensile strength of tablets

of unprocessed material (0 kN/cm) with the tensile strength of granules at the same

compression pressure after dry granulation. Based on a modification of this approach it is

possible to compare the reworkability over the whole compression pressure range (see

6.2.3.1.1, Eq. (33)).

Figure 25 Reworkability index tablets –MacroPactor –

MCC 2:1 LAC/MCC 1:1 LAC, TSratio = tensile strength of tablets of

compacted blend in proportion to unprocessed blend Eq. (33), mean


The reworkability decreased with a higher specific compaction force for both formulations

(see Figure 25). The differences of the reworkability between both formulations at the same

specific compaction force increased from 11.7 % (2 kN/cm) to 14.0 % (8 kN/cm). 8 kN/cm

has to be carefully considered because of the side leakage (see Figure 16). However, the

TSratio showed that the formulation MCC 1:1 LAC had a higher reworkability within the same

scale because of intense fracturing during tableting, insensitivity to MGST, lower susceptible


PROCESS SETTINGS

49

to the specific compaction force (work hardening) and insensitivity to the particle size

enlargement effect. (MiniPactor results can be found in the APPENDIX 8.1.2)

4.2.4.2 Tabletability between different scales

MCC 2:1 LAC - Distinct differences of the tabletability between both scales were found. In

Figure 26 and Figure 27 a comparison of the tabletability between both scales for the same

specific compaction force is depicted (2 and 4 kN/cm, 6 and 8 kN/cm).

Figure 26 Tabletability – MacroPactor/MiniPactor at specific compaction force

of 2 and 4 kN/cm – MCC 2:1 LAC, mean (n = 50), error bars

(standard deviation of mean), scattered lines indicate 10 %

differences to the mean tensile strength of the MacroPactor

A pre-evaluation showed a relative standard deviation of tensile strength at a distinct

compression pressure in the range of 6.72 % (298 MPa) up to 10.00 % (60 MPa) (see

APPENDIX 8.2.2). This was attributed to the relative standard deviation of the compression

pressure at the rotary tablet press (relative standard deviation (RSD) 5 %). Hence, it was

defined that the mean tensile strength of different scales has to be in range of 10 % of the


PROCESS SETTINGS

50

reference to achieve comparable tabletability. Statistical evaluation (T-test) showed

significant differences (p 0.05) between the tablets based on the MacroPactor and

MiniPactor for all values within one specific compaction force except for compression

pressure of 60 MPa at 6 kN/cm and 8 kN/cm (see APPENDIX 8.1.2). At 2, 4 and 6 kN/cm the

MiniPactor’s granules achieved a higher tensile strength compared to the MacroPactor

exceeding the +10 % limit. A smaller difference between the two scales was found for the

specific compaction force 8 kN/cm, but still near +10 %. This was caused by side leakage of

unprocessed material, whereby smaller uncompacted material slipped above the side seal

system, as mentioned at 4.2.3, and induced a higher tensile strength of the tablets for the

MacroPactor.


of 6 and 8 kN/cm – MCC 2:1 LAC, mean (n = 50), error bars



Granules of a roller compaction process have a higher resistance against consolidation during

tableting in comparison to unprocessed material [44]. To illustrate this, the Heckel plot is the


PROCESS SETTINGS

51

most commonly applied technique [106] (see Yield pressure 4.1.1.2.1.1). “In-die” tableting

was performed to measure the Yield pressure of the granules.

Table 6 Heckel - Yield pressure granules - MiniPactor/MacroPactor - MCC 2:1 LAC

Yield

pressure

Py

(1/slope)

0 kN/cm 2 kN/cm 4 kN/cm 6 kN/cm 8 kN/cm

MacroPactor 174.21 176.31 182.45 187.34 186.82

MiniPactor 174.21 177.79 181.09 184.53 185.80

Difference 0 -1.48 1.36 2.81 1.32

n = 3; Py = Yield Pressure Heckel

Results showed that the Yield pressure increased with rising specific compaction force for

both scales. Differences of the Yield pressure between scales were very small ( 2 %). It is

well known that the determination of the Heckel plot is prone to errors in respect to PSD,

surface of the particles and internal lubrication for granules. Granules had a different PSD and

were lubricated using MGST. For these reasons, the results of the Heckel plot have to be

considered carefully [106]. He et al. (2007) also recognized only a small increase of the Yield

pressure for dry granulated MCC at increased specific compaction force because of internal

lubrication, which can disguise the obviously effect of work hardening indicated by the Yield

pressure. For these granules, analysis by Heckel plot is not the method of choice to investigate

the work hardening effect.


PROCESS SETTINGS

52

TSratio gave better insights into the tabletability of both scales compared to Yield pressure.

Reworkability was higher for the granules of the MiniPactor for all specific compaction forces

and showed a higher reworkability. Differences decreased from 16.0 % (2 kN/cm) to 4.6 %

(8 kN/cm) at higher specific compaction forces, which can be positive correlated with the

observed decreased difference of the solid fraction of ribbons between both scales with rising

specific compaction force (see 4.2.2, Figure 13).

Figure 28 Reworkability index tablets - MiniPactor/MacroPactor – MCC 2:1

LAC, TSratio = tensile strength of tablets of compacted blend in

proportion to unprocessed blend Eq. (33), mean (n = 350), error bars


Particle size distribution of the granules is one main factor, which can influence the tensile

strength of tablets. Smaller particles have a greater specific surface to build more interparticle

bonds which can result in a higher tensile strength [45,90]. The differences of particle size

distribution between both scales at same specific compaction force were low (Figure 14,

Figure 17). The d50 showed only small differences and the part of fine particles ( 63 µm)


PROCESS SETTINGS

53

were similar. Thus, the influence of particle size distribution on the tensile strength was

negligible.

The distinguished difference between these granules was the production scale, i.e. compactor

size, which resulted in a different solid fraction of the ribbon. A higher solid fraction at the

MacroPactor was obtained (> 10 %), followed by 10 % lower tensile strength of the tablets

compared to the MiniPactor. A higher solid fraction reflects a higher consolidation and thus a

higher consumption of plastic deformation, which will not be available for tableting (work

hardening effect) again. Roller compaction of formulation MCC 2:1 LAC at different scales

led to a different quality of the tablets although a comparable particle size distribution of the

granules was achieved.


PROCESS SETTINGS

54

MCC 1:1 LAC - Different results were obtained for formulation MCC 1:1 LAC. In Figure 29

tensile strength vs. compression pressure is plotted for four different specific compaction

forces. Considering Figure 29, a relevant difference of the tensile strength of the tablets

between both scales could not be observed.


of 2, 4, 6 and 8 kN/cm – MCC 1:1 LAC, mean (n = 50), error bars



Statistical examination (T-Test) of the tensile strength of tablets at the same compression

pressure between the two scales showed significant differences (p 0.05) for 16 values of 28

(see APPENDIX 8.2.2). However, important to note is that T-Test will detect a significant

difference between both scales when difference of only 5 % will be obtained, assuming a

comparable variance for both samples and previously chosen sample size of n = 50. In

particular, a high compression pressure resulted in a low RSD of the tensile strength (see

exemplary power calculation for 5 % difference APPENDIX 8.2.2). As mentioned above,

10 % difference of tensile strength was defined as acceptable range, so that a relevant


PROCESS SETTINGS

55

difference for low compression pressure would be identified with a power of 0.94. Only two

out of 28 values differed with more than 10 % from the reference i.e. MacroPactor (2 kN/cm

139 MPa, 4 kN/cm 100 MPa). Comparing Figure 26, Figure 27 (MCC 2:1 LAC) and Figure

29 (MCC 1:1 LAC), a more linear course of the tabletability plot was found for MCC 1:1

LAC, which can be correlated to the linear increase of the tabletability of the brittle LAC.

LAC showed a higher impact on the consolidation behaviour with an increased amount in the

formulation (MCC 1:1 LAC). Yield pressure measurements were done but did not allow any

implication, as mentioned above (see APPENDIX 8.1.2).

TSratio of MCC 1:1 LAC showed only a small difference between both scales (max. 3.9 %,

2 kN/cm) compared to MCC 2:1 LAC (max. 16 %, 2 kN/cm, see Figure 28), which confirmed

the impression based on the trend observed in the tabletability plots. The differences between

both scales were low and decreased with rising specific compaction force.

Figure 30 Reworkability index tablets - MacroPactor/MiniPactor–

MCC 1:1 LAC, TSratio = tensile strength of tablets of compacted blend

in proportion to unprocessed blend Eq. (33), mean (n = 350), error

bars (standard deviation of mean)


PROCESS SETTINGS

56

The tensile strength of the tablets of formulation MCC 1:1 LAC is comparable. Regarding the

influence of the particle size on tensile strength, granules of the larger scale exhibited a

similar d50 at low specific compaction forces and a higher d50 at 6 kN/cm and 8 kN/cm

(see Figure 20). Usually it would be expected that the granules of the MiniPactor reaches a

higher tensile strength caused by a smaller PSD, but the high fraction of brittle LAC decreases

the influence of the particle size enlargement effect on tensile strength [107], because the

granules will fracture into smaller particles or even primary particles. This effect was even

more pronounced at higher specific compaction forces (see Figure 30), as the impact of the

brittle LAC increased at high specific compaction forces compared to plastic MCC. Wu et al.

(2007) dry granulated different single brittle excipients, tableted different sieve fraction of the

granules and stated that the granule size had a negligible influence for the tabletability of

brittle granules. A lower amount of MCC (MCC 1:1 LAC) reduced the effect of work

hardening (lower reworkability), followed by a smaller decrease of the tensile strength (TSratio

67.1 %, at 8 kN/cm) caused by specific compaction force compared to MCC 2:1 LAC (TSratio

53.3 %, at 8 kN/cm, see Figure 25). This was enhanced as LAC shows less sensitivity towards

MGST than MCC [49].

The obtained difference in solid fraction of ribbons at same process conditions for the large

scale (> 9 %, see Figure 13) influenced the particle size distribution of granules for the

formulation MCC 1:1 LAC. The expected particle size enlargement effect on tensile strength

was levelled through fracturing of LAC and the lower fraction of MCC in an acceptable range

of 10 % tensile strength between scales. Tabletability of formulation MCC 1:1 LAC at both

scales can be considered as equal. Despite a shift in PSD between Macro- and MiniPactor for

formulation MCC 1:1 LAC, no difference in tabletability between the two scales could be

observed. Obviously, the brittle character of LAC governs the consolidation and bonding

properties under compression, making the formulation less susceptible towards work

hardening and lubricant sensitivity


PROCESS SETTINGS

57

4.2.5 Summary

Two formulations (MCC 2:1 LAC, MCC 1:1 LAC) were dry granulated at two scales (50 mm

roll width MacroPactor, 25 mm roll width MiniPactor) with equal process parameters. A

higher solid fraction of the ribbons was obtained for both formulations at the MacroPactor.

Solid fraction of the ribbons influenced the particle size of the granules in a different way.

Same particle size distribution of granules was achieved for the predominantly plastic

deforming formulation MCC 2:1 LAC, whereby coarser granules were the result for the

predominantly brittle deforming formulation MCC 1:1 LAC at higher specific compaction

force (MacroPactor, 6 kN/cm, 8 kN/cm). Granules of the MacroPactor resulted in a lower

tensile strength of tablets compared to granules of the MiniPactor for formulation MCC 2:1

LAC.

Solid fraction ribbon

Particle size granules

Tensile strength tablets

MCC 2:1 LAC

é

=

ê

MCC 1:1 LAC

é*

=

Roller compaction

Imp

act

Influence of larger scale (MacroPactor)

*6, 8 kN/cm

é

Figure 31 Summary - Impact of formulation and scale, obtained differences of

intermediate products: solid fraction of ribbons, particle size of

granules and tensile strength of tablets caused by larger scale

(MacroPactor) compared to smaller scale (MiniPactor), equal sign =

equal, arrow up/down = higher/lower, MCC = Microcrystalline

cellulose, LAC = α - Lactosemonohydrate


PROCESS SETTINGS

58

Thus, the influence of the higher solid fraction corresponded directly to the tensile strength of

the tablets, because of the high amount of MCC, which is mainly influenced by specific

compaction force (work hardening) and sensitivity towards MGST. Only small differences of

the tensile strength between both scales were observed for formulation MCC 1:1 LAC,

although particle size distribution differed. This was driven by the impact of the brittle LAC,

which resulted in negligible susceptibility towards specific compaction force (work

hardening), MGST and particle size enlargement effect in respect to tensile strength of tablets.

The second formulation MCC 1:1 LAC showed a higher robustness towards scalability. The

impact of a different solid fraction at different scales on the tablet quality will be increased

due to a higher amount of MCC in the formulation (work hardening effect).

Differences between the two scales in downstream processing (solid fraction, granules,

tablets) have not been described before in literature. It can be concluded that using the same

process settings for identically constructed roller compactors of different scale will not

necessarily lead to the same quality of tablets, because the quality is formulation and scale

dependent. The solid fraction of the ribbon was confirmed as an essential quality aspect in

respect of a successful scale up. Differences of the solid fraction between scales were

observed in this chapter and will be further investigated and explained in the following

chapters.

ADAPTED PROCESS SETTINGS OF DIFFERENT SCALES TO ACHIEVE SIMILAR

PRODUCT QUALITY

59

4.3 ADAPTED PROCESS SETTINGS OF DIFFERENT SCALES TO

ACHIEVE SIMILAR PRODUCT QUALITY

Formulation MCC 2:1 LAC was dry granulated using the MiniPactor at various specific

compaction forces to achieve the target solid fractions of the MacroPactor, in order to reduce

the difference between scales with respect to the tensile strength of tablets (see 4.2).

Throughout this manuscript, the results of the adapted process parameter at the MiniPactor are

called Scale Model.

4.3.1 Achieving the same solid fraction of ribbon by using adapted process

parameter settings

Data of the previously measured solid fraction of the Macro- and MiniPactor at four specific

compaction forces were used to determine the specific compaction forces (see Figure 13),

which were required to achieve equal solid fraction for the MiniPactor. For this purpose,

specific compaction forces were plotted versus solid fractions of the ribbons. Existing data of

the solid fractions of the MiniPactor were fitted by a polynomial equation to obtain the

required specific compaction forces (Scale Model).


PRODUCT QUALITY

60

Figure 32 Scale Model approach – Adapting specific compaction force of

MiniPactor to achieve target solid fraction of ribbon (MacroPactor) by

polynomic fitting, mean (n = 5), error bars (standard deviation of

mean)

Target solid fractions of the MacroPactor were set in this correlation function to predict the

required specific compaction force at the MiniPactor (4.1, 7.0, 9.6 and 11.4 kN/cm, see Figure

32). The predicted specific compaction forces were used for processing. Subsequently, the

solid fraction of the ribbons was measured to verify the proposed Scale Model. Adapted

specific compaction forces for the Scale Model are reported as 2, 4, 6 and 8 kN/cm.

53.923x²-39.03+7.578

R²= 0.9995


PRODUCT QUALITY

61

Figure 33 Solid fraction ribbon – Results Scale Model (adapted process

parameter settings) and MiniPactor (equal process parameters), mean

(n = 5), error bars (standard deviation of mean), scattered lines

indicate 5 % range to the target solid fraction (MacroPactor), SF =

Solid fraction

Results of the solid fraction are depicted in Figure 33. T-test (p 0.05) results showed

differences between three of four solid fractions between target solid fraction and the solid

fraction of the Scale Model (see APPENDIX 8.2.3). Only the solid fraction for the specific

compaction force of 7.0 kN/cm (0.7086 SF) showed no difference to the target solid fraction.

However, the differences between both scales (applying adapted process parameter settings)

were only -0.12 % to 4.39 %, compared to 10.20 % to 17.90 % by using equal process

parameter settings (see Figure 13). All solid fractions of the ribbons were in a 5 % range to

the target solid fraction. Results on downstream processing and an evaluation of this strategy

for different scales will be discussed in the next sections. Based on previous findings, it can

be assumed that the solid fraction is the most influential factor on downstream processing


PRODUCT QUALITY

62

(see 4.2) and therefore that a similar granule and tablet quality will be achieved at a

comparable solid fraction of the ribbon at different scales.

4.3.2 Particle size distribution and porosity of granules

PSD - Application of the Scale Model resulted in a higher d50 for the whole range of

investigated specific compaction forces (see Figure 34), which is in accordance to results in

4.2.3.2 where identical force settings resulted in a similar PSD at a lower SF for granules

produced with the MiniPactor. T-test (p 0.05) confirmed a difference between all granules

at same solid fraction (see APPENDIX 8.2.3).

Figure 34 d50 granules - MacroPactor/MiniPactor/Scale Model, mean (n = 3),

error bars (standard deviation of mean)

As mentioned above, higher specific compaction forces resulted in coarser granules at the

same roller compactor. Analysing the PSD confirmed that coarser granules were obtained for

the Scale Model (see Figure 35). Comparing all scales, a lower amount of fine particles was

found for the Scale Model (see Figure 35). The impact of the milling process of the granulator

on PSD between both scales could be demonstrated to be negligible (see 4.2.3.3). A potential


PRODUCT QUALITY

63

impact factor and explanation between scales can be the solid fraction distribution of the

ribbon along the roll width, which will be discussed in the next section (see 4.4).

Figure 35 PSD granules– MacroPactor/MiniPactor/Scale Model at specific

compaction force of 2, 4, 6 and 8 kN/cm, triangle (mean, n = 3), error

bars (standard deviation of mean)

Porosity & Appearance – Measurements of the intragranular porosity by mercury

porosimetry enable a deeper insight of differences between these granules. Other authors have

shown that the particle size influences the measurement [110,111]. Therefore, sieve fractions

of the granules were used (90, 180, 250 µm) in order to avoid measuring the interparticular

voids (bimodal pore size distribution, see Figure 66), and to ensure a good comparability

between granules of MacroPactor, MiniPactor and Scale Model. Unprocessed blend of

MCC 2:1 LAC had a d50 of 103 µm. Sieve fractions of 180 µm and 250 µm will define

agglomerated particles and 90 µm the fine particles. All granules were sieved and separated.

Cumulative sum of the intruded mercury volume over a pore range of 0.4 – 1.8 µm (see

6.2.1.4) was used as porosity index. A linear decrease of the porosity was observed at

increased solid fraction of the ribbons (see Figure 36, left plot). Solid fraction of ribbon


PRODUCT QUALITY

64

directly correlated with the porosity of granules in a linear coherence. Linear regression

showed a good correlation coefficient (R2) of 0.8950. Granules generated by the MacroPactor

(SF = 0.8102, 8 kN/cm) were excluded for linear regression, due to the aforementioned side

seal leakage. For this reason unprocessed material slipped above the compaction zone (see

Figure 16), which resulted in a higher porosity (grey circles, left plot).

Figure 36 Granule porosity – MCC 2:1 LAC - Left plot: Cumulative sum

relative intruded mercury volume [mm³/g] (pore radius range 0.4 -

1.8 µm) vs. solid fraction ribbon – Right plot: Cumulative sum

relative intruded mercury volume [mm³/g] (pore radius range 0.4 -

1.8 µm) vs. sieve fraction of 90, 180, 250 µm, grey circles = excluded

for linear regression (MaroPactor at 8 kN/cm see 4.2.3.1)

Herting et al. (2008) have shown a decrease of the surface area of a dry granulated MCC sieve

fraction with increased specific compaction force, which was correlated to a lower porosity.

Nordstrom et al. (2015) have demonstrated a decrease of the porosity of re-tableted milled

tablets of MCC for a granule sieve fraction of 500 – 710 µm with increasing compression

pressure, but both did not evaluate the relation between the porosity of granules and the solid


PRODUCT QUALITY

65

fraction of ribbons. This evaluation of a decreasing porosity for granules with increasing solid

fraction seems to be more appropriate for dry granulated granules, because a negative linear

correlation was determined (see Figure 36, left plot). A high solid fraction correlated to a low

porosity of granules. It was assumed that a coarser sieve fraction of 250 µm would be strained

with a higher load of pressure at the same specific compaction force compared to a fine sieve

fraction of 90 µm. The reason is that more particles were agglomerated under pressure and

this results in a denser granule with lower porosity. Comparing the porosity of the sieve

fraction 90, 180, 250 µm within the MacroPactor, MiniPactor or Scale Model at 2, 4, 6 or

8 kN/cm, no tendency of the porosity could be observed (see Figure 36, right plot). This can

be attributed to the influence of the complex milling process by the granulator, whereby the

particles are mainly sheared and sliced of the ribbons against the sieve mesh [112,113].


PRODUCT QUALITY

66

Figure 37 Appearance of particles at 4 kN/cm - Sieve fractions 90, 180, 250 µm

of MacroPactor/MiniPactor/Scale Model

Particle morphology of the three sieve fractions within one specific compaction force and

scale was visually inspected by microscopy (see Figure 37). The larger sieve fractions of

180 µm and 250 µm appeared rougher as they contained particles of a higher degree of

agglomeration compared to the smaller sieve fraction of 90 µm. Considering the particle

morphology, no differences between the MacroPactor, MiniPactor and Scale Model granules`

sieve fractions were observed (see Figure 37).

Scale Model

MiniPactor

MacroPactor


PRODUCT QUALITY

67

A comparison of the porosity of the sieve fractions between different scales is depicted in

Figure 38. For better illustration, the porosity of granules were normalized to the reference

value (MacroPactor) for each sieve fraction; Porosities of the MiniPactor and Scale Model

were subtracted from this reference value (see Figure 38). A positive value indicates a lower

porosity and negative value a higher porosity compared to the MacroPactor.

Figure 38 Granule porosity – Residual values cumulative sum relative intruded

mercury volume [mm³/g] (pore radius range 0.4 -1.8 µm) to target

granule porosity (MacroPactor) - Sieve fraction of 90, 180, 250 µm

Compared to the MacroPactor, the MiniPactor showed a lower solid fraction of the ribbon and

the Scale Model achieved a similar solid fraction. The solid fraction of the ribbon can be

correlated with the porosity of the granules. Granules of the MiniPactor led to higher

porosities, in particular at the specific compaction force of 4 kN/cm and 6 kN/cm, whereby

the porosities of the Scale Model were always closer to the reference value i.e. MacroPactor

(see Figure 38). The specific compaction force at 2 kN/cm had a minor influence on the

differences on the porosities of the granules, because the impact of the solid fraction of the

ribbon was less. The first weak bonding between the particles occurred at a low solid fraction.

Higher porosity

Lower porosity

Reference line

(MacroPactor)


PRODUCT QUALITY

68

Considering all specific compaction forces, a small tendency of the granules of the Scale

Model to lower porosities compared to the MacroPactor were observed. A different result was

obtained for the 8 kN/cm, for which the Scale Model showed a difference to the reference

value, while the MiniPactor had a similar porosity, which was caused by the slipped

uncompacted material of the MacroPactor at 8 kN/cm (see 4.2.3.1). It can be concluded that a

same solid fraction of the ribbon (MacroPactor & Scale Model) led to similar porosity of the

granules. Therefore, the next step was to investigate how the different particle size

distributions and a similar porosity will affect the tabletability and compressibility of the

granules.


PRODUCT QUALITY

69

4.3.3 Tabletability and compressibility influenced by material attributes of

granules and ribbons

Same process conditions were applied for tableting of the granules of the Scale Model

compared to the granules of the MacroPactor and MiniPactor. No practical relevant difference

between the tensile strength of MacroPactor and Scale Model was observed (< 10 %). T-Test

between MacroPactor and Scale Model showed a significant difference (p 0.05) for most

tablets (see APPENDIX 8.2.3). An example calculation for the power of the T-Test indicated

that the T-Test would detect a 5 % difference of the tensile strength as significant with a

power up to 0.97 (see APPENDIX 8.2.3).

Figure 39 Tabletability –MacroPactor/MiniPactor/Scale Model at specific

compaction force of 2, 4, 6 and 8 kN/cm, mean (n = 50), error bars



A pre-evaluation showed a relative standard deviation of 10 % for the tensile strength within

one compression pressure for the same scale (see 4.2.4). Hence, a 10 % deviation of the


PRODUCT QUALITY

70

tensile strength is an acceptable range between two scales and the difference between

MacroPactor and Scale Model is negligible. Considering the tabletability of granules of the

Scale Model, these showed same tensile strength for the 2 kN/cm, 4 kN/cm and 6 kN/cm (see

Figure 39). The resulting tensile strength was in a very good accordance and had a tensile

strength lower than the desired acceptable deviation of 10 % to tablets of the MacroPactor

(see 4.2.4). The achieved tensile strength of the applied 8 kN/cm was lower (see explanation

4.2.3 & 4.2.4).

Obviously, the coarser granules of the Scale Model (particle size enlargement effect) had a

minor impact on the tensile strength and did not lead to a lower tensile strength of the tablets.

The agglomerated particles i.e. granules collapsed into smaller particles during tableting.

The particle size enlargement effect caused by an initial coarser PSD of the Scale Model

becomes negligible. The impact of coarser granules will only gain more influence if the

granule strength is increased, which will prevent a fracturing of agglomerated particles

[25,50]. Nordstrom et al. (2015) [25] have shown that the porosity of the dry granulated

granules of MCC essentially influenced the tensile strength of tablets. Dry granulated granules

with a high porosity will collapse into smaller particles during compression to tablets

followed by a closer “intergranular void structure of a tablet” (e.g. solid fraction) [25] and an

increased tensile strength. Additionally, the fracturing of granules into smaller particles can be

enhanced if the mixture contains a brittle component [10,48] (see Figure 25). Effects like the

porosity of granules and the more pronounced fracturing caused by the amount of brittle LAC

in the formulation, diminished the effect of the coarser granules for the Scale Model. A

confirmation of a low impact of the PSD on the tensile strength for this formulation (MCC 2:1

LAC) was observed for the MiniPactor granules. A similar PSD was found compared to the

MacroPactor granules, which resulted in a higher tensile strength of the tablets for the

MiniPactor (see Figure 39), reflecting the achieved lower solid fraction of the ribbon (see

Figure 33) and the higher porosity of the granules i.e. fracturing tendency (see Figure 36).


PRODUCT QUALITY

71

The behaviour of the fracturing and densification process of granules can be evaluated by

considering a compressibility plot. A different fracturing behaviour under pressure will result

in a different microstructure (e.g. solid fraction) of the tablet at the same compression

pressure level. Thereby, granules with a high fracturing or consolidation tendency will be

more compressible. This will lead to a higher solid fraction at the same compression pressure

compared to granules, which will be more “resistant to deformation of tableting” (work

hardening) [50].

Figure 40 Compressibility –MacroPactor/MiniPactor/Scale Model at 2, 4, 6 and

8 kN/cm, mean (n = 50), error bars (standard deviation of mean)

The compressibility plot of the tablets indicated that the MiniPactor achieved a higher solid

fraction for all investigated compression pressures (see Figure 40). This was caused by the

higher porosity and fracturing tendency of the granules of the MiniPactor. The granules of the

Scale Model demonstrated a similar compressibility compared to the MacroPactor. The

compressibility plot confirmed that the initial coarser granules of the Scale Model had no

influence on the compressibility and proved that the porosity of the granules levelled the

granule size enlargement effect (lower bonding surface). The MacroPactor granules of


PRODUCT QUALITY

72

8 kN/cm confirmed this assumption as they gained a higher compressibility, which was

caused due to a higher amount of unprocessed material with a higher porosity (see Figure 16).

Figure 41 Reworkability index tablets –MiniPactor/MacroPactor/Scale Model,

TSratio = tensile strength of tablets of compacted blend in proportion to

unprocessed blend Eq. (33), mean (n = 350), error bars (standard

deviation of mean)

Tensile strength of the tablets of the Scale Model was in good accordance with the tensile

strength of the MacroPactor (see Figure 41). The proposed Scale Model resulted in a

successful scale up regarding tabletability and compressibility. However, it needs to be further

investigated why different ribbon solid fraction was obtained at identical process parameter at

two scales (see 4.4).


PRODUCT QUALITY

73

4.3.4 Summary

It was demonstrated that same process settings at two different scales for a roller compaction

process did not result in an equal tensile strength of the tablets. Therefore, a new approach

was proposed (Scale Model) to first predict and then adapt the specific compaction force of

the smaller scale (MiniPactor) to achieve the same target solid fraction of the ribbons of the

larger scale (MacroPactor).

Ribbons were produced with the desired target solid fraction in a range of 5 % in

comparison to ribbons produced with the MacroPactor.

Solid fraction ribbon

Particle size granules

Porosity granules

Tensile strength tablets

Equal process parameter

settings

ê

=

é

é

Adaption of specific

compaction force

é

=

=

=

Roller compaction

Imp

act

Imp

act

Figure 42 Summary – Impact of scale for formulation MCC 2:1 LAC

(MCC = Microcrystalline cellulose, LAC = α –Lactosemonohydrate),

differences of intermediate products: solid fraction of ribbons, particle

size of granules, porosity of granules and tensile strength of tablets of

the small scale (MiniPactor) at equal process parameters and adapted

specific compaction force (Scale Model) compared to target

intermediate quality attributes of the large scale (MacroPactor),

equal sign = equal, arrow up/down = higher/lower


PRODUCT QUALITY

74

The resulting tensile strength of the tablets of the Scale Model was well in accordance with

the target tensile strength of the tablets of the MacroPactor. The porosity of the granules

demonstrated a negative linear correlation with increased solid fraction of the ribbon.

Although the same solid fraction between both scales was achieved (MacroPactor, Scale

Model), the resulting PSD distribution of the granules was different. The roll width is one

impact factor, resulting in a different solid fraction distribution along the ribbon width, which

will be investigated in the next section (see 4.4). However, the effect of granule size

enlargement (lower bonding surface, Scale Model) was negligible as a similar granule

porosity was achieved compared to the granules of the MacroPactor, which caused an equal

“resistant to deformation of tableting” [50] (work hardening), a comparable fracturing

behaviour (see Figure 40) and indicated a similar granule strength [25]. The impact of the

fracturing tendency of the granules as an important factor for the tensile strength was proved

in the previous chapter 4.2 for formulation MCC 1:1 LAC. It is common knowledge that a

coarser granule size distribution (lower bonding surface, Scale Model) leads to a different

compressibility and tensile strength, because of a lower surface area of the particles compared

to finer particles. In contrast, this was not shown for this formulation (MCC 2:1 LAC) and

process parameters as it was demonstrated that the resulted coarser particle size distribution of

the adapted process parameters led to a similar tensile strength of the tablets.

A simplified summary of the effects and results for formulation MCC 2:1 LAC MiniPactor

compared to the MacroPactor is provided in Figure 42. The Scale Model approach

demonstrated a practicable solution to achieve a successful scale up during process

development for a predominantly plastic deforming formulation. Nevertheless, a higher

specific compaction force at the small scale (MiniPactor) was required to achieve the same

solid fraction compared to the larger scale (MacroPactor). This aspect will be in investigated

in the next chapter with a newly developed analysis method to better characterise the ribbons.

INVESTIGATION OF SOLID FRACTION DISTRIBUTION ALONG THE ROLL WIDTH

BETWEEN DIFFERENT SCALES VIA NIR AT-LINE

75

4.4 INVESTIGATION OF SOLID FRACTION DISTRIBUTION ALONG THE

ROLL WIDTH BETWEEN DIFFERENT SCALES VIA NIR AT-LINE

In the previous chapters it was demonstrated that ribbons produced with either the MiniPactor

or the MacroPactor applying identical process parameters showed different solid fraction.

These obtained difference will be investigated with a near infrared reflectance method to

characterise the solid fraction of ribbons along the roll width. Different authors have reported

about a solid fraction distribution along the roll width within one scale [33,40,114–118]. This

solid fraction distribution will be investigated with an NIR method between scales

(MacroPactor, MiniPactor). NIR measurement allows measuring smaller parts of the ribbon in

comparison to the GeoPycnometer method; this method was limited by the sample size (three

pieces of 8 mm* 8 mm, see 6.2.2.3). Furthermore, applying the at-line NIR analysis results in

a relevant time saving. Afterwards the previously proposed approach for a scale up will be

evaluated by NIR in respect to the solid fraction. A formulation containing 21 % Metformin

(w/w) with MCC 2:1 LAC will be used as model active pharmaceutical ingredient (API).



76

4.4.1 Method evaluation to determine solid fraction along the roll width by

GeoPycnometer and NIR

As reference method for the NIR spectra (see Figure 43) the solid fraction was analysed using

a GeoPycnometer. It was a prerequisite to measure small pieces of the ribbon with the

reference method, because the NIR probe acquires a small area on the ribbon of about

0.5 cm².

Figure 43 Schematic drawing - Process flow development NIR method

An appropriate and reproducible analytical method applying the GeoPycnometer will only be

achieved if the sample size (ribbon) fills 20 % of the whole vessel volume (DryFlo® + ribbon

volume). A pre-evaluation showed that a lower limit of the reference method was reached by

analysing at least three cut pieces of 8 mm* 8 mm from the respective ribbon. Reference

method for NIR was performed with a vessel of 12.7 mm diameter and a plunger force of

28 N. Sample volume in the vessel was always about 20 %. Ribbons were taken after gap

settings reached steady state conditions (see 3.1). A small grid was scratched on the ribbon to

identify the sample squares. Spectra of the squares were acquired by NIR. Afterwards, the

ribbon was cut and separated into centre and edges (front, back) to also analyse the same

ribbon with the GeoPycnometer. Spectra of three squares were correlated to one reference

value for calibration purpose. A schematic flow process is depicted in Figure 43.



77

4.4.1.1 GeoPycnometer as reference method – Solid fraction along the roll width

A sample ribbon of 6 kN/cm (MiniPactor) was measured with the GeoPycnometer to evaluate

whether differences occur along the roll width. A distance-weighted least square fitting was

performed between measured samples of a ribbon (GeoPyc) in order to illustrate the

difference between measured points.

Figure 44 Evaluation solid fraction distribution of ribbon along roll width

measured by GeoPycnometer– Exemplary ribbon produced with

6 kN/cm, black circles = reference value by GeoPycnometer, colour

indicates different solid fractions

Figure 44 indicates a decrease of the solid fraction from the centre to the edges of the ribbon.

This finding is in agreement with literature, as various authors have demonstrated a higher

solid fraction at the centre of the ribbon for a cheek plate side seal system

[21,30,31,36,40,41,115]. A non-uniform distribution of the solid fraction along the roll width

is caused by the side seal system [33,115]. Using cheek plates will lead to friction force

between powder and plates, whereby the powder entry into the gap is reduced at the edges.

This side seal friction force counteracts the friction force between rolls and powder [116].

Therefore, a heterogeneous solid fraction can be observed. Reference method

(GeoPycnometer) detected differences along the roll width and was appropriate for the

purpose of an NIR method development.

Solid fraction



78

4.4.1.2 NIR spectra of a ribbon along the roll width

Multiple exemplary NIR spectra of a ribbon compacted at 6 kN/cm are depicted in Figure 45.

A difference between locations of the ribbon was detected by the NIR, which were indicated

by a baseline shift of the spectra. This is consistent with the literature. Various authors

observed a higher absorption of NIR spectra with an increased pressure for tablets (i.e.

increased solid fraction) [119–121] and ribbons [32,38,122].

Figure 45 NIR spectra of a ribbon at 6 kN/cm – Front = blue, middle = red,

back = green

A higher absorption was positively correlated to a high solid fraction caused by less

reflectance of the radiation in diffuse reflectance mode (see 6.2.2.4). Applying the above-

mentioned NIR methodology differences along the roll width could be identified. Hence, this

technique was considered as suitable technique to identify differences between scales.



79

4.4.1.2.1 Factors influencing the spectra of the ribbon

The overall target when developing a new analytical method is to ensure high accuracy,

precision and robustness. It is essential to investigate different factors impacting these

characteristics and how a proper measurement setup or data pre-process techniques reduce

this interference [123]. Measurement setups were evaluated to ensure these characteristics.

Different factors were considered and evaluated to find a reliable setup:

1. Sample condition (time after processing, damaged surface )

2. Measuring setup (distance probe to sample, angle probe to sample, light in the room)

3. Software setup (resolution of scans)

All measurements were done applying the NIR diffuse reflectance mode at a wavenumber

range of 12.000 – 4.000 cm-1.

4.4.1.2.1.1 Sample condition

All samples were stored for 48h at 0 – 65 % relative humidity and 19 - 25°C in respect to their

elastic recovery after roller compaction. The smooth side of the ribbon was analysed as it was

easier to scratch the grid on the ribbon. Surface of the ribbon can be irregular or damaged

during processing, e.g. caused by powder adhesion on surface of the rolls or abrupt

segmentations of the ribbon.

Figure 46 Absorption ribbon surface – Blue = damaged, red = normal



80

A ribbon sample was damaged for illustration by a knife to simulate a different surface. The

square was measured before and after this manipulation. A higher absorption of the damaged

surface could be observed because of the irregular shape of the surface, which enhanced the

scattering [124], so that less radiation reached the detector and resulted in an offset (see

Figure 46). Derogation of this offset can be obtained with data pre-processing techniques (see

6.2.2.4.2). First derivation levelled the observed baseline shift and will contribute to a reliable

NIR-method (see Figure 47).

Figure 47 Surface: First derivation – Blue = damaged, red = normal



81

4.4.1.2.1.2 Measuring setup

Distance - Different distances of 0, 3, 6 and 12 mm from the probe to the sample were

evaluated. Calibrated metal blocks clamped between the probe to adjust the position.

Figure 48 Distance probe to sample – Red = 0 mm, blue = 3 mm, green = 6 mm,

pink = 12 mm

Ten spectra of the same sample were acquired for each distance. A higher absorption with

increased distance and a rising variation within the same distance between the 10 spectra were

obtained (see Figure 48). A higher distance led to less detection of the radiation by the

detector and resulted in a higher absorption.



82

Considering the first derivation of the spectra, a noisy signal could be observed with increased

distance to the sample caused by a higher detection of stray light (see Figure 49). The distance

of 0 mm was chosen as the best distance to the sample, which showed less variability between

different spectra and a low noise signal. The zero position was defined as the longest part of

the probe touching the sample.

Figure 49 Distance probe to sample: First derivation - Red = 0 mm, blue =

3 mm, green = 6 mm, pink = 12 mm



83

Angle –Two angles of probe to the sample were evaluated. The probe had a bevel of 20°.

Therefore, a 70° and 90° angle were examined. At 70° the whole surface of the probe was in

contact with the sample. Ten re-adjustments of the zero position probe to sample within one

angle were done. As depicted in Figure 50, an angle of 70° led to a higher variability within

one setup. At an angle of 70° the laser beam of the probe was no longer apparent. This made

it more difficult to adjust the position and find the same position on the sample square. Based

on this a 90° angle was set as measuring setup.

Figure 50 Angle probe to sample – Red = 70°, blue = 90°

90°

70°



84

Light – The light of the room may influence the absorption level. An evaluation with light

and without light did not show any impact on the absorption. All spectra coincided. As it was

easier to recognize the laser beam of the probe without light in the room and adjust the

position on the ribbon, it was decided to use no light in the room. First derivation showed no

difference between both (not shown).

Figure 51 Light – Pink = off, blue = on



85

4.4.1.2.1.3 Software setup resolution

As a next step the software setup was evaluated. Different resolutions lead to a smoothing of

the spectra over 8, 16 or 32 wavenumbers. Thus, a higher resolution indicates a higher

smoothing and more information about the spectra gets lost. Figure 52 shows the first

derivation of ten spectra for every resolution. A low resolution of 8 led to a noisy signal

especially at higher wavenumbers. A resolution of 32 diminished too much spectral

information. A resolution of 16 was chosen as software setup. The number of scans did not

show any impact on the spectral information, a number of 64 scans was used.

Figure 52 Resolution: First derivation - Red = 8, blue = 16, pink = 32



86

4.4.1.2.1.4 Specificity – Impact of composition on NIR spectra – Definition

spectral range

NIR spectra can be used to derive chemical information. In particular by spectra peaks, which

are representing characteristic chemical bonds of the used formulation (see 6.2.2.4). An

adaption of the spectral range is necessary dependent on the formulation. It was a prerequisite

for the measurement of the solid fraction (physical property) by NIR to diminish this impact

in order to develop a reliable method, which is independent of the fraction of API or

excipient. NIR spectra of the pure API, excipients and blend were acquired to evaluate an

appropriate spectral range for the physical property of the ribbon.

Figure 53 Spectra components and blend - Black = Metformin hydrochloride,

green = − Lactosemonohydrate, red = blend, blue = Sodium

carboxymethylcellulose, pink = Microcrystalline cellulose

Different peaks occurred depending on the chemical structure of the components (see Figure

53). Specific peak maxima were observed for Metformin (combination band 5000 – 4000 cm-

1, first overtone band about 6570 cm-1, second overtone band about 9596 cm-1) [125] and

water (combination band about 5200 cm-1, overtone band about 7000 cm-1) [126] (see

APPENDIX 8.1.3). Based on these results, an appreciable spectral range above 9597 cm-1 will

be evaluated for method development. At this spectral range, the impact of the composition

showed less impact compared to a range of lower wavenumbers.

N-H

N-H2

N-H N=NH

H

Water -OH

Defined spectral

range



87

4.4.1.2.1.5 Summary

A summary of the evaluated NIR settings is depicted in Table 7. Using the first derivation of

the spectra will reduce the impact of the measuring setup and sample condition.

Table 7 Summary - Evaluation NIR setup

Factor Setup Defined

Sample condition

Storage time 48 h (relative humidity 0-65 %,

temperature 19-25°C)

Ribbon surface Smooth

Measuring setup

Distance probe to sample 0 mm

Angle probe to sample 90°

Ambient light Off

Software setup

Resolution 16

Scans 32

Spectral range No influence of composition above 9597 cm-1



88

4.4.2 Development of an at-line NIR method for solid fraction measurements of

ribbons

4.4.2.1 Principal component analysis (PCA) – Verify spectral range

Multivariate data analysis was applied to describe the variance of the spectra over the range of

wavenumbers (11941 cm-1 – 9665 cm-1). Evaluation of an appropriate spectral range can be

done by principal component analysis (PCA) (see 6.2.2.4.1). PCA verifies if the observed

variance of the spectra can be distinguishing the solid fractions. A vast amount of data is

compressed to principal components representing the variance of the spectra.

Figure 54 Principal component analysis – Ribbon at 6 kN/cm – Spectral range

above 9597 cm-1, location on the ribbon: front = blue, middle = red,

back = green

Principal component 1 showed promising result as spectra were clustered by solid fraction. A

distinction can be made by front, back and centre of the ribbon. The second principal

component did not account for any difference between solid fractions. Principal component

two was equally distributed over the spectra. Defined spectral range was appropriate to

acquire a bigger sample size and to proceed with internal (cross validation) and external

validation (test set).

0.66 – 0.68 Solid fraction

(front, back)

(

0.74– 0.77

Solid

fraction

(middle)

Principal component 2

Pri

nci

pal

com

ponen

t 1



89

4.4.2.2 Calibration data set – NIR method development

The calibration data set consisted of ribbons of the MiniPactor compacted at 4, 6, 8 and 10

kN/cm. In literature various authors correlated the slope of a linear regression over the whole

spectral range [29,32,38,127] in order to determine the observed baseline shift (see 4.4.1.2)

with increasing solid fraction. As investigated, various factors will have an impact on this

baseline shift (e.g. distance probe to sample, see 4.4.1.2.1). First derivation allows to level

these baseline shifts resulting in a higher accuracy, precision and robustness. Therefore, the

first derivation was used for all spectra (see 4.4.1.2.1).



90

4.4.2.3 Results - Internal validation – Cross validation

Full cross validation of the calibration data set was performed by multivariate statistical

method of partial least square analysis (PLS) (see 6.2.2.4.1). Each sample was removed from

the full data set and predicted based on the regression results for the remaining samples (leave

one out method).

Figure 55 Results internal validation –

Predicted vs. observed solid fraction ribbon, colour indicates different

solid fraction ranges

A good correlation between predicted and observed values was found (see Figure 55). The

correlation coefficient was 0.8680. Five principal components were necessary to reach this

correlation coefficient (rank, PLS). Compared to the PCA, more principal components were

necessary to describe this bigger data set. Furthermore, the root mean square error of cross

validation and bias were low, which indicated a reliable model (see Table 8).



91

Table 8 Summary – Results internal validation

No. of calibration spectra included 699

Calibration range [Solid fraction] 0.5853 – 0.8066

Reference measurements [GeoPycnometer] 233

Data pre-processing technique 1st derivation

Seleceted wavenumber range 11941 cm-1 – 9658 cm-1

Coefficient of correlation (R2) 0.8680

Root mean square error of cross validation

(RMSECV) 0.0188

Rank (PLS-factor) 5

Residual prediction deviation (RPD) 2.81

Bias -0.0000366



92

4.4.2.4 Results - External validation – Test set validation

Unknown ribbon samples of 4, 6, 8 and 10 kN/cm were taken to evaluate the performance of

the developed method. Additionally, three drug loads (DL) of 15 %, 21 % (target drug load)

and 27 % were embedded in validation set to prove independence of the drug load.

Figure 56 Results external validation – Left plot: Predicted vs. observed solid

fraction ribbon, colour indicates different solid fraction ranges of the

ribbon – Right plot: Predicted vs. observed solid fraction ribbon,

colour indicates different drug loads (%) for Metformin

Results of external validation were well in agreement with cross validation results. Similar

correlation coefficients (0.8611, 0.8680) and a low root mean square error of prediction and

bias could be found. (see Figure 56, left plot, Table 9). For the investigated drug loads of the

ribbon an impact on the accuracy of predicted solid fraction over the investigated range of

solid fractions was not found (see Figure 56, right plot). Predicted solid fractions were evenly

distributed around the observed solid fractions (GeoPycnometer). No trend in the residuals

could be identified for the different drug loads (see Figure 56, right plot). Validation results

confirmed that predictions of the solid fraction of unknown samples were successful. An



93

appropriate method was developed to further proceed with a scale up approach and to

investigate the distribution of the solid fraction along roll width at different scales.

Table 9 Summary - Results external validation

No. of dosage strengths 21 %, 15 %, 27 %

No. of validation spectra included 216

Calibration range [Solid fraction] 0.5853 – 0.8066

Reference measurements [GeoPycnometer] 72

Data pre-processing technique 1st derivation

Seleceted wavenumber range 11941 cm-1 – 9658 cm-1

Coefficient of correlation (R2) 0.8611

Root mean square error of prediction

(RMSEP) 0.0189

Rank (PLS-factor) 5

Residual prediction deviation (RPD) 2.68

Bias -0.00425



94

4.4.3 Scale up approach – Comparison of solid fractions of ribbons of a Metformin

formulation at two scales

The blend was compacted with the MiniPactor and MacroPactor at 4, 6, 8 and 10 kN/cm.

Results of the solid fraction measurements are shown in Figure 57 (GeoPycnometer).

Differences of the solid fraction of the ribbon were obtained between the two scales.

MacroPactor achieved a higher solid fraction compared to the MiniPactor (Figure 57, left

plot). These results were consistent with the results presented in previous chapters (see 4.2,

4.3).

Figure 57 Comparison solid fraction ribbon MacroPactor/MiniPactor/Scale up -

Left plot: solid fraction ribbon vs. specific compaction force - Right

plot: specific compaction force vs. solid fraction ribbon, mean (n = 5),

error bars (standard deviation of mean), DL = drug load, Met =

Metformin

T-Test indicated differences (p 0.05) in SF between the MacroPactor and MiniPactor at all

specific compaction forces (see APPENDIX 8.2.4). Solid fraction at 6 kN/cm of the

Polynomic fitting

99.926*x^2

-106.38*x+30.353

R²= 0.9989



95

MiniPactor was defined as target solid fraction (0.7063) for a scale up (MacroPactor), which

would be an appropriate value for the solid fraction of a ribbon to gain an acceptable product

quality [17]. A polynomic fitting was done (see Figure 57, right plot), as previously described

in chapter 4.3, to adapt the specific compaction force of the MacroPactor, and to achieve the

target solid fraction of the MiniPactor. The polynomic equation allowed calculating the

required specific compaction force at the MacroPactor by employing the target solid fraction

of 0.7063 (MiniPactor, see Figure 57, right plot). Solid fraction of the adapted specific

compaction force of 5.1 kN/cm (MacroPactor, black triangle) was 0.7040 and showed only a

difference of -0.3 % to the target solid fraction. Ribbons were taken for NIR measurements.



96

4.4.4 Solid fraction distribution along the roll width between different scales by

NIR

Independent samples of all specific compaction forces were analysed applying the NIR

method. In Figure 58 the average solid fraction along the ribbon width for each scale at

different specific compaction forces is shown. A polynomic fitting was done to illustrate the

differences between the ribbons produced at two scales, where 0 cm represents the middle of

the ribbon.

Figure 58 Mean solid fraction along ribbon width at compaction forces of 4, 6, 8

and 10 kN/cm, Scale up 5.1 kN/cm, mean (n = 12), error bars

(standard deviation of mean), ribbon length 9.20 cm

Solid fraction increased from the edges to the centre from about 0.61 to 0.70 (+15 %) for both

scales due to cheek plates effects (see 4.4.1.1) [21,30,31,36,40,41,115]. Ribbons of the

MacroPactor obtained a narrow solid fraction distribution along the roll width because the

impact of the cheek plates became lower with a larger distance to the edges (see 4.4.1.1). The

ribbons produced with the MacroPactor had a larger area with a high solid fraction, which

explains the higher “total” solid fraction measured by the GeoPycnometer for the ribbons of



97

the MacroPactor. This was previously observed with equal process parameters for both scales

(see 4.2.2). For process development purposes, it is interesting to note that the standard

deviation of the solid fraction increased with increasing specific compaction force (see Figure

58). At a higher specific compaction force, a higher amount of powder has to be dragged into

the compaction zone by the tamp auger in order to reach steady state conditions at a gap of

e.g. 3 mm. This can only be realized if the speed of the tamp auger is increased. Therefore, the

impact of the periodical rotation by the tamp auger is enhanced, which is followed by a more

diverse solid fraction along the ribbon length compared to lower specific compaction forces.

Figure 59 Solid fraction distribution ribbon (MacroPactor 6 kN/cm) -

Left plot: solid fraction along ribbon length (contour plot, fit type

= distance weighted least square), triangle measurement points of NIR

- Right plot: solid fraction distribution of ribbon along ribbon width,

triangle measurement points of NIR, circle (mean, n = 12), error bars


A sinusoidal periodical course of the solid fraction along the ribbon length, attributed to the

periodically screw rotation of the tamp auger [29,36,39], could be exemplary shown for a

ribbon of the MacroPactor (see Figure 59).



98

Additionally, the results of solid fraction distribution explained the lower observed tensile

strength of tablets by the MacroPactor compared to the MiniPactor (see 4.2.4.2). A high solid

fraction will lead to higher consumption of plastic deformation, lower porosity of granules

and thus to a lower tensile strength of tablets (see 4.3.4). Considering the scale up approach,

the sample showed an equal “total” solid fraction compared to the MiniPactor (see Figure 57).

This comparable solid fraction will result in similar granule and tablet properties (see 4.3.2 &

4.3.3). A lower solid fraction can be obtained by adapting the specific compaction force

(Scale up 5.1 kN/cm) to counteract the different solid fraction distribution along the roll width

between two scales. Thus, adapting the specific compaction force by measurements of the

“total” solid fraction (GeoPycnometer) is a suitable scale up strategy for a roller compaction

process. A uniform distribution of the solid fraction at different scales would be preferable,

but due to construction issues like roll width and side seal system, difficult to gain.



99

4.4.5 Summary

In previous chapters a different solid fraction of the ribbon at equal process parameters at

different scales was obtained. A more efficient NIR method was developed to measure the

solid fraction of ribbons along the roll width, as the GeoPycnometer method was limited by

sample size. A new formulation with a model API Metformin was used. Various factors were

investigated, which have a potential impact on the reliability of the NIR method. Sample

conditions, measuring setup and API content. Afterwards internal validation and external

validation was performed with independent samples, which illustrated a good correlation

coefficient R2 0.86 to the reference method. It was possible to predict the solid fraction of

unknown samples by acquiring the NIR spectra comprising reduction of analysis time.

The developed NIR method allowed measuring the solid fraction distribution along the full

roll width. A difference between the solid fraction of the ribbons based on different scales was

found. MacroPactor showed a larger area with a high solid fraction along the roll width

compared to the MiniPactor. The effect of the cheek plates (lower solid fraction at the edges),

decreased with increased distance to the cheek plates ( 1 cm, see Figure 58), which was

especially the case for the MacroPactor with a broader roll width. This leads to a higher

“total” solid fraction of the ribbons produced by the large scale compared to a small scale at

equal process parameters. These results explained different quality attributes of granules and

tensile strength of the tablets impacted by the solid fraction of ribbons, which were observed

in previous chapters at equal process parameters. The proposed scale up approach showed that

the differences of resulting granules and tablets between scales can be balanced through

adaption of the specific compaction force. Furthermore, this approach can be used in the

pharmaceutical industry during process development from small development batches to

commercial batches to satisfy market demands.



100

MODEL DEVELOPMENT – PREDICTING SOLID FRACTION OF A TABLET

101

4.5 MODEL DEVELOPMENT – PREDICTING SOLID FRACTION OF A

TABLET

Previous chapters have shown that compression behaviour of mixtures (solid fraction vs.

compression) have an impact on roller compaction process. Thus, it would be beneficial to

predict the compression behaviour of mixtures based on compression analysis of single

excipients to support decision guidance for formulation development purposes.

The predictive power of the Percolation theory, the Kawakita model, and a simple exponential

model were systematically evaluated for different direct compression formulations. Four

mixtures were compressed over a wide pressure range at various fractions of microcrystalline

cellulose (MCC) and pre-agglomerated lactose monohydrate (LAC). Formulations contained

different amounts of MCC and LAC in range of 72 % to 24 % (see Table 17). First, all models

were applied to single components and thereafter an additive rule was used (see 3.3.1.4), for

predicting the solid fraction of the mixtures, reflecting the composition of the formulation.

Finally, the prediction and the observed solid fractions of the compressed mixtures were

compared to evaluate the suitability of the models (see Figure 60).

Figure 60 Process flow - Prediction solid fraction of tablets


102

4.5.1 Application of models – Excipients

4.5.1.1 Input data for model application

A low number of input values are a prerequisite to ease applicability of a model. Thus, solid

fraction after compression and true density measurements (see Table 2) of MCC, LAC and

CARB were used as input parameters. Forty-two tablets were compacted between 50 MPa –

350 MPa for each excipient. The tablets were weighed directly after ejection with an

analytical balance. The diameter and the height of each tablet were determined by an

automatic tablet tester to enable calculation of the solid fraction according to Eq. (34) and Eq.

(35). Tablets only consisting of LAC compressed at a compression pressure of 50 MPa were

too weak for being measured in the tablet tester. Results of the compressibility are depicted in

Figure 61.

Figure 61 Excipients - Solid fraction vs. compression pressure [MPa], mean

(n= 6), error bars (standard deviation of mean)


103

True densities of mixtures were calculated according to Eq. (17). Results can be seen in

Table 10.

Table 10 Calculated true density mixtures

Mixture #1

MCC 3:1 LAC

Mixture #2

MCC 2:1 LAC

Mixture #3

MCC 1:1 LAC

Mixture #4

MCC 1:3 LAC

1.5540 [g/cm3] 1.5527 [g/cm3] 1.5504 [g/cm3] 1.5468 [g/cm3]

MCC = Microcrystalline cellulose; LAC = − Lactosemonohydrate

4.5.1.2 Percolation

The different parameters in the model can be interpreted with regard to compressibility. The

maximum achievable value of SF/SFmax is limited by the value of SFmax and cannot exceed the

value of 1 (see Figure 62).

Figure 62 Excipients - Percolation SF/SFmax vs. compression pressure


104

SFmax of MCC and LAC had a similar value which caused a similar value of 𝑆 [128] although

they had a different course of their compressibility graph. In contrast to the plastic deformable

materials MCC and CARB, LAC reached a plateau of SF exceeding a compression pressure

of 300 MPa, while MCC and CARB reached that plateau at pressure values of about

200 MPa. This behaviour is due to the brittle deformation characteristics of LAC [129] which

resulted in lower compressibility compared to plastic deformation, the course seemed to be

steadily increasing which could be a consequence of the pre-agglomerated LAC and induces a

brittle fracture (see Figure 61).

All excipients showed similar percolation thresholds, where the percolation threshold of LAC

was lower than the threshold of MCC. As the percolation threshold is defined as the pressure

limit above which particle rearrangement has completed and an infinite cluster formed (see

3.3.1.1), excipients of higher bulk or working density will need less pressure to settle, i.e.

rearrange. The working density (bulk density) can be derived from 𝑉0-values at 1-2 MPa (see

3.3.1.2 & 4.5.1.3) and were 0.9376 g/cm³ (𝑉0 = 0.2117 cm³) for LAC and 0.4996 g/cm³

( 𝑉0 = 0.3990 cm³) for MCC which explains the lower percolation threshold for the

spherically shaped Tablettose 80® (LAC) particles in contrast to the fibrous Avicel PH 102®

(MCC) particles.

Table 11 Estimated parameters for the Percolation model

Modell

Parameter MCC LAC CARB

𝑆𝐹

𝑆𝐹𝑚𝑎𝑥

= 𝑆 ∗ (𝑥𝑝 − 𝑝𝑐)𝑞

𝑜𝑟 𝑆𝐹 = 𝑆 ∗ (𝑥𝑝 − 𝑝𝑐)𝑞

∗ 𝑆𝐹𝑚𝑎𝑥

SFmax (measured) 0.9264 0.9242 0.8338

Estimate Estimate Estimate

S 0.7248 0.7183 0.6778

𝑝𝑐 48.7851 46.4764 48.3177

q 0.05615 0.05565 0.07033

R² 0.9753 0.9306 0.9715

MCC = Microcrystalline cellulose; LAC = − Lactosemonohydrate; CARB = Sodium carboxymethylcellulose;

SF = Solid fraction; SFmax = Maximal solid fraction; S = Proportional constant; xp = Compression pressure

[MPa]; pc = Percolation threshold; q = compressibility exponent; R2 = correlation coefficient


105

The compressibility exponent q provided a reasonable interpretation of the course of the

graphs (see Figure 62) and relates to the steepness of the compressibility plot approximating

𝑆𝐹𝑚𝑎𝑥 (maximal densification) over the whole range of the applied compression pressure.

CARB showed the highest value and achieved the maximal densification already at 200 MPa,

whereas MCC and LAC exhibited similar values of q, which was driven by the behaviour

discussed earlier (particle shape) and can also be explained by considering Eq. (6), which

caused normalisation through maximum measured solid fraction (SFmax). The coefficient of

correlation was 0.9306 and showed good fitting results for the model (see Table 11).

4.5.1.3 Kawakita

According to the modification of Kawakita C (see 3.3.1.2), a proof for the new definition of 𝐶

is exemplarily shown for microcrystalline cellulose in Figure 63 where C = V0 – VP/ V0 is

correlated to solid fraction (SF). V0 was determined in-die for microcrystalline cellulose

(n = 3) to be in a typical range of 1-2 MPa. The linear correlation between both definitions

throughout the range of interest (for tableting & roller compaction) for solid fraction from

0.72 - 0.91 was confirmed by the high coefficient of correlation (R2 0.9951) (see Figure 63).

LAC and CARB showed similar behaviour resulting in correlation coefficients of 0.9954 and

0.9928 respectively (see APPENDIX 8.1.4).

Figure 63 “In-die” data MCC- Kawakita C vs. derived C equal to solid fraction

Considering the applicability of the Kawakita model, the linearity of the plot P/SF vs

compression pressure is a specified condition. For all excipients, good linearity was


106

demonstrated and therefore a high coefficient of correlation could be achieved (R² 0.9992)

over the whole range of compression pressure (see Table 12, Figure 64).

Figure 64 Kawakita - Excipients P/SF vs. compression pressure [MPa]

This proves that it was possible to determine Vmin by a Helium-Pycnometer and to

consequently use solid fraction as 𝐶 . Thus, the Kawakita parameter 𝑎 can be seen as

maximum strain or by the definition of Eq. (8) as maximum achievable solid fraction.

Kawakita 𝑎 decreased in the order of MCC > LAC > CARB, which was consistent with the

observed values of SFmax determined for the Percolation model shown above (see Table 11),

and emphasised that MCC was the most compressible excipient. Considering b-1, MCC

showed the highest deformation capacity (low value), followed by LAC and CARB.


107

Table 12 Estimated parameters for the Kawakita model

Modell

Parameter MCC LAC CARB

Kawakita a 0.9614 0.9522 0.8836

Kawakita b.-1 17.0257 17.7252 24.7369

Fitted

Parameter

MCC LAC CARB

Estimate Estimate Estimate

Slope 1.0401 1.0502 1.1317

Intercept 16.3693 16.8779 21.8582

R2 0.9992 0.9992 0.9992

MCC = Microcrystalline cellulose; LAC = − Lactose monohydrate; CARB = Sodium carboxymethylcellulose;

R2 = correlation coefficient

4.5.1.4 Exponential

Results of parameter estimation of the exponential model showed a good coefficient of

correlation R² 0.9282. For LAC the R² was quite low because LAC had a more linear course

of the graph, as mentioned above (see Table 13).

Table 13 Estimated parameters for the Exponential model

Modell

Parameter

Estimate

MCC LAC CARB

d 0.9153 0.9342 0.8335

f -0.5257 -0.1928 -0.5345

g -0.0179 -0.0062 -0.0172

R2 0.9916 0.9282 0.9960


d, f, g = Exponential constants; R2 = correlation coefficient


108

4.5.2 Prediction of solid fraction – Mixtures

All three models were applied to the excipients and then multiplied by their mass fraction

according to the composition of the mixture to predict the solid fraction Eq. (16) (see 3.3.1.4).

In order to compare the prediction with the observation, the mean value of six tablets was

used. For MCC 1:3 LAC at 50 MPa no reasonable values could be determined because the

tablets were too soft.

Figure 65 Results models – Plot A: Predicted vs. observed solid fraction –

Plot B: Sum least squares of model for each formulation and total sum

least squares of the models – Plot C: Residuals vs. compression

pressure range [MPa]

For the exponential model a slight trend of overpredicting the solid fraction can be seen (see

Figure 65A). However, according to this the results of the three models differ in terms of the

observed values only in a range of -0.017 to 0.030, which is only a small difference for solid

fraction. Considering the absolute residuals a relevant deviation over the compression

pressure cannot be anticipated. The total sum of the least squares (see Figure 65B) follows the

order Percolation < Kawakita < Exponential model, which demonstrates that the Percolation

A B

C


109

model yields the best prediction of the solid fraction for all mixtures. The exponential model

has a stronger overprediction of the solid fraction, whereby the Percolation leads to a

moderate underprediction (see Figure 65C). Verifying the model’s quality of prediction by

subdividing the sum of least squares by mixtures, gives an enhanced ability for interpretation

(see Figure 65B). For the mixtures #1 - #3, which had a higher fraction of MCC, the

percolation model achieved a much more precise prediction than Kawakita or the Exponential

model. A reason for the good prediction of the solid fraction by the percolation model can be

that it also takes into account the maximum achievable solid fraction (SFmax), which provides

a limit for the system. The poor prediction by the percolation model for the mixture #4

MCC 1:3 LAC can be caused by the lower correlation coefficient 0.9306 for LAC (see Table

11), whereas Kawakita has a correlation coefficient of 0.9992 (see Table 12). Nevertheless

Mixture #4 has a rather untypical high load of LAC (72.37 %) for a tablet formulation. This

could be an interesting point for further investigations to verify if the percolation model is

also in good agreement for formulations, which have similar brittle compression behaviour

like LAC. Despite this the Kawakita model gives the best coefficient of correlation (see Table

12) for all excipients, it resulted in valuable predictions of all mixtures. Comparing

predictions of solid fraction by Kawakita to recent papers, which focus on the prediction of

porosity by Kawakita [75,79,81], residuals in a similar range from -0.39 % to 3.12 %

compared to Busignies et al. (2012) were found , lower than 2.5 % [79], which is in a good

agreement. Also Frenning et al. (2009) have shown that it is possible to predict the effective

Kawakita parameters a and b-1 of spherical pellets out of single compression analyses.

However, this reflects that it is possible to derive C as solid fraction (see 3.3.1.2), which is

defined by measured Vmin and Vp, to get reasonable results for the prediction of solid fraction

by Kawakita and to obtain a possible comparison between various research laboratories.


110

4.5.3 Summary

A modified Kawakita model and the Percolation theory were successfully adapted for the

compressibility plot to predict the solid fraction of four compositions using single

compression analysis of the excipients. These two mechanistic models, were both compared

with a statistical model based on a simple exponential function. Both theoretical models

showed very good prediction of solid fraction. The Percolation model seems to be more

suitable for a common composition of tablets in pharmaceutical industry, whereas Kawakita

showed better results for mixtures which deform predominantly by brittle fragmentation. The

limited input data set (tablet geometry/mass, true density) and common software for

regression analysis, makes it easy to establish a more complex dataset for different excipients

providing systematic guidance for the formulator. Further investigations are needed to

consider mixtures with different active pharmaceutical ingredients and excipients with brittle

fracture.

In respect to roller compaction process, Percolation and Kawakita can provide an opportunity

to predict the solid fraction of ribbons if an correlation between compression pressure at the

tablet press and specific compaction force of the roller compactor will be established, which

will reduce the number experiments at a roller compactor to find the right process settings for

a specified formulation.

MATERIALS

111

5. MATERIALS

5.1 MATERIALS

Table 14 Materials

Name Trade name Supplier Batch #

−Lactosemonohydrate (LAC) Tablettose 80®

Meggle Wasserburg GmbH & Co.

KG, Megglestraße 6-12, 83512

Wasserburg am Inn, Germany

203403

200292

203403

Microcrystalline cellulose (MCC) Avicel® PH102 FMC Biopolymer Co, 1735 Market

St., Philadelphia, PA 19103, USA

400994

905814

903926

Sodium carboxymethylcellulose

(CARB) AcDiSol®

FMC Biopolymer Co, 1735 Market

St., Philadelphia, PA 19103, USA

401433

408157

802637

Magnesium stearate

(MGST) LIGAMED®

Peter Greven GmbH & Co. KG, Peter-

Greven-Straße 20-30, 53902 Bad

Münstereifel, Germany

304668

308990-1

Metformin hydrochloride (MET) Metformin

hydrochloride

Weifa AS, Gruvevn. 1, P.O. Box 98,

NO-3791 Kragera, Norway

401889

Graphite powder DryFlo®

Micromeritics Instrument

Corporation, 4356 Communications

Dr, Norcross, GA 30093, USA

-

−Lactosemonohydrate (LAC) is a commonly diluent in tablets. −Lactosemonohydrate

occurs by a crystallisation of a supersaturated solution of lactose below 93°C [91]. A

processed product is Tablettose 80®, whereby fine milled lactose particles agglomerate

during a continous spray drying processs by water. The resulting agglomerates show

increased flowability and increased tabletability, due to eased fracturing into smaller particles

compared to single crystals of the size of the agglomerates. LAC is an excipient, which shows

a typical brittle compression behaviour [90,91].

Microcrystalline cellulose (MCC) is a partially depolymerized cellulose derived from

−cellulose wood pulps using sulfuric acid, followed by a washing step and a spray drying

MATERIALS

112

process, which results in fibrous MCC particles [130]. MCC is one of the frequently used

excipients as diluent and binder, because of its good plastic deformation under pressure [60].

Sodium carboxymethylcellulose (CARB) is a cross-linked polymer of carboxymethylcellulose

sodium used as disintegrant as it has a high water uptake ability combined with an high

swelling property to provide a fast disintegration of the tablet.

Magnesium stearate (MGST) is a lubricant to reduce the sticking or friction between stainless

steel (e.g. rolls, punches) and powder. Metformin hydrochloride (MET) is an oral

antihyperglycemic drug, used for type 2 diabetes. The API belongs to the biguanide class.

Metformin has been used as model API in chapter 4.4. Different fractions of LAC and MCC

were employed in this thesis. CARB and MGST were on constant levels (see 5.2).

FORMULATIONS

113

5.2 FORMULATIONS

5.2.1 Placebo composition - Chapter 4.1, 4.2 and 4.3

Table 15 Placebo composition – Chapter 4.1, 4.2 and 4.3

MCC 2:1 LAC MCC 1:1 LAC

Component Content % (w/w) Content % (w/w)

Microcrystalline cellulose 64.0 48.0

−Lactosemonohydrate 32.0 48.0

Sodium carboxymethylcellulose 3.0 3.0

Magnesium stearate 1.0 1.0


5.2.2 API composition - Chapter 4.4

Table 16 API composition (Metformin) – Chapter 4.4

DL 21 %

(Target DL) DL 27 % DL 15 %

Component Content % (w/w) Content % (w/w) Content % (w/w)

Metformin hydrochloride 21.0 27.0 15.0

Microcrystalline cellulose 50.0 46.0 54.0

−Lactosemonohydrate 25.0 23.0 27.0

Sodium

carboxymethylcellulose 3.0 3.0 3.0

Magnesium stearate 1.0 1.0 1.0

DL = Drug load

FORMULATIONS

114

5.2.3 Placebo composition- Chapter 4.5

Table 17 Placebo composition – Chapter 4.5

Mixture #1

MCC 3:1 LAC

#2

MCC 2:1 LAC

#3

MCC 1:1 LAC

#4

MCC 1:3 LAC

Component Content % (w/w) Content % (w/w) Content % (w/w) Content % (w/w)

Microcrystalline

cellulose 72.36 64.32 48.24 24.13

−Lactosemonohydrate 24.12 32.16 48.24 72.37

Sodium

carboxymethylcellulose 3.02 3.01 3.02 3.00

Magnesium stearate 0.50 0.50 0.50 0.50


MANUFACTURING AND TECHNOLOGIES

115

6. MANUFACTURING & ANALYTICAL METHODS

6.1 MANUFACTURING AND TECHNOLOGIES

Table 18 Overview experimental methods

Chapter Equipment Parameter Material attribute

4.1 Material

characterisation -

True density, Particle size distribution,

Bulk/Tapped density, Tabletability,

Compressibility, Compactibility

4.1 MacroPactor 2, 4, 6, 8 kN/cm

Solid fraction ribbon:

Throughput method, Mercury porosimetry,

GeoPycnometer (GeoPyc)

4.2 MacroPactor /

MiniPactor 2, 4, 6, 8 kN/cm

Particle size distribution, Solid fraction ribbon,

Tabletability, Compressibility

4.3 MiniPactor 4.1, 7.0, 9.6, 11.4

kN/cm

Solid fraction ribbon, Particle size distribution,

Particle appearance, Porosity granules,

Tabletability, Compressibility

4.4 MacroPactor /

MiniPactor

4, 5.1, 6, 8, 10

kN/cm

Solid fraction, Solid fraction distribution along

roll width/ribbon length with near infrared

reflectance

0 FlexiTab 50 – 350 MPa Solid fraction tablet


116

Table 19 Manufacturing equipment

Name Process Manufacturer

Comil 196 U Pre-Screening Quadro Engineering Corp., Canada

Handscreen Screening Kressner, Germany

Tumbling blender Blending Servolift GmbH, Germany

Turbula blender Blending Turbula type 2A, Willy A. Bachofen

AG Maschinenfabrik, Germany

MiniPactor Dry granulation Gerteis Maschinen + Process-

engineering, Switzerland

M1075-GMP-Polygran® Walzenpresse Dry granulation Gerteis Maschinen + Process-

engineering, Switzerland

Fette 1200 Tableting Fette Compacting GmbH, Germany

Fette 1200i Tableting Fette Compacting GmbH, Germany

FlexiTab Tableting Röltgen GmbH & Co. KG, Germany


117

6.1.1 PREPARATION OF MIXTURES AND PROCESS FLOW CHARTS

6.1.1.1 Mixture for roller compaction

Table 20 Manufacturing flow chart: Mixture

Step Equipment Materials Operation In-process

controls

1.0 Screening mill Microcrystalline cellulose

→

Pre-screening

Bulk/tapped

density, PSD

− Lactose monohydrate


Metformin hydrochloride

↓

2.0 Freefall blender Pre-blending

↓

Pre-blend

3.0 Handscreen Magnesium stearate → Screening

↓

Handmix Equal weight parts of Pre-blend (Step

2.0) → Hand mix

↓

Hand screen-mix 0

4.0 Freefall blender Pre-blend (Step 2.0)

→ Blending

Handscreen-mix 0 (Step 3.0)

↓

Dry-mix

Bulk/tapped

density,

Hausner ratio,

PSD


118

Table 21 Process parameters: Mixture

Step Process step Description of manufacturing process Process parameter

1.0 Pre-screening

Microcrystalline cellulose, −Lactose monohydrate,

Sodium carboxymethylcellulose and Metformin are

prescreened with Comil 196U

Mesh size : 1.016 [mm]

Rotations: 630 [rpm]

Sieve: rasp

2.0 Pre-blending Pre-screened is blended in freefall container blender

Time: 20:00 [min]

Rotations: 10 [rpm]

Bin: 1000L (190L

Metformin)

3.0 Screening

Magnesium stearate is screened with a handscreen

and then manually mixed with equal weight parts of

Pre-blend

Mesh size: 1.000 [mm]

4.0 Blending Pre-blend and Handscreen-mix is blended in freefall

container blender

Time: 10:00 [min]

Rotations: 10 [rpm]

Bin: 1000L (190L

Metformin)

Table 22 Batch size: Mixture

Composition Name Batch size [kg] Total [kg]

Placebo

MCC 2:1 LAC 248.75

497.5

MCC 1:1 LAC 248.75

Metformin

MET 21 67.66

73.66 27 % DL 3

15 % DL 3


119

6.1.1.2 Chapter 4.2 and 4.3

Table 23 Manufacturing flow chart: Chapter 4.2 and 4.3


controls

6.A Roller compactor Dry-mix (Step 4.0) → Dry compacting Solid fraction

ribbon

↓

Milling

↓

Granules Bulk/tapped

density, PSD

7.A Handscreen Magnesium stearate → Screening

↓

Handmix Magnesium stearate with equal weight

parts of granules (Step 6.A) → Hand mix

↓

Hand screen-mix A

8.A Freefall blender

Granules (Step 6.A)

→ Final blending

Handscreen-mix A (Step 7.A)

↓

Final blend

9.A Rotary tablet press Final blend (Step 8.A) → Tableting

↓

Tablet cores

Tensile

strength, solid

fraction


120

Table 24 Process parameters: Chapter 4.2 and 4.3


6.A Dry compacting &

milling

Dry-mix is compacted and the ribbons are directly

milled by the roller compactor M1075-GMP-

Polygran® and MiniPactor®

Ratio tamping auger/feeding

auger: 160 [%]

Roll speed: 2 [rpm]

Automatic gap control: ON

Gap size: 3 [mm]

Granulator speed: 70 [rpm]

Granulator angle: 360°/360°

[cw/ccw]

Sieve: 0.8 [mm], square

Milling: Star rotor

Chapter 4.2:

Compaction force: 2, 4, 6, 8

[kN/cm]

Chapter 4.3 (Scale Model):

Compaction force: 4.1, 7.0,

9.6, 11.4 [kN/cm]

7.A Screening



granules


8.A Final blending Granules and handscreen-mix A are blended in

freefall container blender

Time: 10:00 [min]

Rotations: 10 [rpm]

Bin: 40L

9.A Tableting Final blend is compressed into tablet cores with a

rotary press

Compression force: 3, 5, 7, 9,

11, 13, 15 [kN], RSD 5 %

Pre-compression force: 0

[kN]

Compression speed: 80.000

[tbl/h]

Punch size: 8 [mm], round

convex punches, bevelled

edges

Target weight: 220 [mg]


121

Table 25 Batch size: Chapter 4.2 and 4.3


Placebo

MCC 2:1 LAC 10 (each kN/cm)

240 MCC 1:1 LAC 10 (each kN/cm)

Scale Model 10 (each kN/cm)

Table 26 Manufacturing flow chart: Chapter 4.2: Comparison milling process


controls

6.B Single punch press Dry-mix (Step 4.0) → Tableting

↓

Tablet cores Solid fraction

7.B Roller compactor Tablet cores (Step 6.B) → Milling

↓

Granules PSD


122

Table 27 Process parameters: Chapter 4.2: Comparison milling process


6.B Tableting Dry-mix is compressed into tablet cores with a single

punch press

Compression force: is

defined by target SF

Target SF :

MCC 2:1 LAC : 0.62 (6.5

kN), 0.77 (16.0 kN)

MCC 1:1 LAC: 0.64 (6.7

kN), 0.79 (18.0 kN)


and biplane


Feeder: ON

Automatic lubrication: press

chamber coating system ON,

every ten tablets

7.B Milling

Tablets are milled within one hour by roller

compactor´s granulator M1075-GMP-Polygran® and

MiniPactor®


[cw/ccw]


Milling: Star rotor

Table 28 Batch size: Comparison milling process


Placebo

MCC 2:1 LAC 1.2 (each SF, 2000

tablets)

4.8

MCC 1:1 LAC 1.2 (each SF, 2000

tablets)


123

Table 29 Manufacturing flow chart: Chapter 4.3: Porosity granules


controls

6.C Sieving Granules (Step 6.A) → Sieving

↓

Sieve fractions granules

(90µm, 180µm, 250µm)

Mercury

porosimetry

Table 30 Process parameters: Chapter 4.3: Porosity granules


6.C Sieving Granules are separated by sieve analysis

Time: 10 min

Amplitude: 2 mm

Interval time: 10 s

Sieve fraction of 90µm,

180µm, 250µm


124

6.1.1.3 Chapter 4.4

Table 31 Manufacturing flow chart: Chapter 4.4


controls

6.D Roller compactor Dry-mix (Step 4.0)

(Metformin composition) → Dry compacting

Solid fraction

ribbon

(GeoPyc, NIR)

Table 32 Process Parameters Chapter 4.4


6.D Dry compacting &

milling

Dry-mix is dry compacted and the ribbons directly

milled by the roller compactor M1075-GMP-

Polygran® and MiniPactor®

Ratio tamping auger/fedding

auger: 160 [%]

Roll speed: 2 [rpm]

Automatic gap control: ON

Gap size: 3 [mm]

Granulator speed: 70 [rpm]


[cw/ccw]


Milling: Star rotor

Specific compaction force:

4, 6, 8, 10 [kN/cm]

Scale up:

Large scale: 5.1 [kN/cm]

Table 33 Batch size: Chapter 4.4


Placebo

MET 21 5 (each kN/cm)

51 27 % DL 3

15 % DL 3


125

6.1.1.4 Chapter 4.5

Table 34 Manufacturing flow chart: Chapter 4.5


controls

I Turbula blender Microcrystalline cellulose

→

Blending − Lactose monohydrate


↓

Pre-blend

II Handscreen Magnesium stearate → Screening

↓

Handmix Equal weight parts of Pre-blend (Step I) → Hand mix

↓

Hand screen-mix

III Freefall blender Pre-blend (Step I)

→ Blending

Handscreen-mix (Step II)

↓

Dry-mix

IV Single punch press Dry-mix (Step III) → Tableting

↓

Tablet cores Solid fraction


126

Table 35 Process parameters: Chapter 4.5


I Pre-blending Material is blended in turbula blender

Time: 03:00 [min]

Rotations: 50 [rpm]

Bin: 2L

II Screening



Pre-blend


III Blending Pre-blend and handscreen-mix is blended in a turbula

blender

Time: 05:00 [min]

Rotations: 50 [rpm]

Bin: 2L

IV Tableting Dry-mix is compressed into tablet cores with a single

punch press

Compression force: 4, 8, 12,

16, 20, 24, 28 [kN]


and biplane


Table 36 Batch size: Chapter 4.5


Placebo

MCC 3:1 LAC 0.5

2

MCC 2:1 LAC 0.5

MCC 1:1 LAC 0.5

MCC 1:3 LAC 0.5

ANALYTICAL METHODS

127

6.2 ANALYTICAL METHODS

Table 37 Analytical equipment

Name Sample Attribute Manufacturer

AccuPyc II Raw material, Unprocessed

blends True density


Corporation, USA

Stampf-Volumeter PT-TD1

+ Engelsmann STAV II Raw material, granules Bulk/Tapped density

Stampf-Volumeter PT-

TD1, Pharma Test

Apparatebau GmbH,

Germany

Retsch AS 200 control Raw material, granules Particle size distribution Retsch GmbH, Germany

Pascal 140-240/440 Granules, ribbons Porosity (mercury), Solid

fraction

Thermo Fisher Scientific,

USA

Keyence VHX 5000 (VH-

Z20R/Z20T) Granules Mircroscopy

Keyence Deutschland

GmbH, Germany

GeoPycnometer

(GeoPyc1360) Ribbons Solid fraction


Corporation, USA

Bruker Multi-Purpose

Analyzer (MPA), Fourier

Transform,

Detector RT-InGaAs

Ribbons Near infrared spectroscopy Bruker Optics, Germany

Micrometer screw Ribbons Micrometer screw MIB – Messzeuge GmbH,

Germany

AT400 Tablets Weight Mettler-Toledo GmbH,

Germany

Erweka TBH 310 MD Tablets Geometry tablets, weight,

breaking force Erweka GmbH, Germany

FlexiTab Raw material, Unprocessed

blends, granules Compression analysis

Röltgen GmbH & Co. KG,

Germany

Erweka TBH 310 MD Tablets Geometry tablets, breaking

force Erweka GmbH, Germany

Kraemer UTS 12 F Tablets Geometry tablets, weight,

breaking force

Kraemer Elektronik

GmbH, Germany

ANALYTICAL METHODS

128

Table 38 Software and data processing

Data Name

In die data (Compressibility) Data Acquisition System DAQ4, Dr. M. Hucke

General data evaluation and figures Spotfire Version 7.5.1.8, Tibco® Software Inc., USA

Polynom fitting Excel, Microsoft® Office Professional Plus 2010

Power calculation G*Power Statistical Power Analyses Version 3.1.9.2

Contour plots and figures Statistica Version 12®, StatSoft GmbH, Germany

T-Test, ANOVA Statistica Version 12®, StatSoft GmbH, Germany

Spectra and near infrared reflectance method

development

OPUS version 7.2®, Bruker Optik GmbH, Germany

Data fitting Chapter 4.5 OriginPro 8G®, OriginLab Corporation, USA

6.2.1 Raw material and granules

6.2.1.1 True density

True density of all excipients and blends was measured by a helium gas pycnometer

(AccuPyc II, Micromeritics Instrument Corporation, USA), the helium penetrated into the

voids between particles or intragranular voids. The displaced volume determined solids’

volume. A blank measurement of the chamber was previously performed (9.2 cm³).

Afterwards metal spheres calibrated the amount of displaced helium. Temperature was

controlled to guarantee same analysis conditions. Fill pressure was set to 19.5 psig with an

equilibration rate of 0.005 psig/min. 10 fill purges resulted in one mean value for the true

density.

6.2.1.2 Bulk/Tapped density

Bulk and tapped density of powders and granules measured in a 250 ml cylinder (Engelsmann

STAV II). A mass of 100 g was filled into the cylinder. The volume for the bulk density was

determined. Afterwards taps of 10, 500, 1250, 2500 were used (Stampf-Volumeter PT-TD1)

to determine the tapped volume. Bulk, tapped density and Hausner ratio were calculated

according to United States Pharmacopeia (USP) [16].

ANALYTICAL METHODS

129

6.2.1.3 Particle size distribution

Particle size distribution of powders and granules were performed in triplicate by sieve

analysis. Sieve analyses was performed with a sieve tower, which comprised sieves of 63, 90,

125, 180, 250, 355, 500, 630, 710 and 1000 µm mesh size. Analysis time was 10 min with an

amplitude of 2 mm and an interval time of 10 seconds (Retsch AS 200 control). Retained

mass on each sieve was determined by an analytical mass balance. Sample size was 100 g

(n = 3). The total amount passed [%] of the sample was calculated for each sieve and plotted

versus the mesh size for a detailed analyses. d50 represents the cumulative mean particle size

calculated by Retsch AS 200 according to USP [16].

6.2.1.4 Mercury porosimetry

Mercury intrusion-porosimetry is the method of choice to measure the porosity of solids.

Mercury is a non-wetting-liquid for the most solids because it has a high surface tension and a

high contact angle of over 90° on solids [131]. Without an external pressure mercury cannot

intrude into pores. Washburn (1921) [132] derived an equation, where he stated that a

capillary has a cylindrical pore. The force of the capillary which prevents the intrusion of the

liquid is dependent of the surface tension of the liquid, the angle of contact with the solid and

the length of the area of contact between liquid and the surface of solid, which can be defined

by 2*π*radius of the pore.

𝐶𝑜𝑢𝑛𝑡𝑒𝑟𝑎𝑐𝑡𝑖𝑛𝑔 𝑓𝑜𝑟𝑐𝑒 𝑐𝑎𝑝𝑖𝑙𝑙𝑎𝑟𝑦 = 2 ∗ 𝜋 ∗ 𝑟 ∗ 𝑦 ∗ 𝑐𝑜𝑠 (𝜃) Eq. (18)

The force required to enter the pore can be defined as:

𝐸𝑥𝑡𝑒𝑟𝑛𝑎𝑙 𝑃 = 𝐹

𝑎𝑟𝑒𝑎 Eq. (19)

𝐸𝑥𝑡𝑒𝑟𝑛𝑎𝑙 𝐹 = 𝜋 ∗ 𝑟2 ∗ 𝑃 Eq. (20)

If are both in equilibrium Eq. (19) and Eq. (20)

2 ∗ 𝜋 ∗ 𝑟 ∗ 𝑦 ∗ 𝑐𝑜𝑠 (𝜃) = 𝜋 ∗ 𝑟² ∗ 𝑃 Eq. (21)

This shows that pore radius is inversely proportional to the applied pressure

𝑟 = −2∗𝑦∗𝑐𝑜𝑠(𝜃)

𝑃 Eq. (22)

r = radius, y = surface tension (liquid), cos (θ) = angle [°] of contact, F = Force, P = Pressure

A capacitive sensor measured the intruded volume of mercury during increasing pressure,

whereby the intruded volume of mercury per sample (relative volume mm³/g) correlates with

pore size distribution.

ANALYTICAL METHODS

130

𝑅𝑒𝑙𝑎𝑡𝑖𝑣𝑒 𝑣𝑜𝑙𝑢𝑚𝑒 (𝑚𝑚3/𝑔) = 𝐼𝑛𝑡𝑟𝑢𝑑𝑒𝑑 𝑚𝑒𝑟𝑐𝑢𝑟𝑦 𝑣𝑜𝑙𝑢𝑚𝑒 𝑎𝑡 𝑟𝑎𝑑𝑖𝑢𝑠 (𝑚𝑚3)

𝑆𝑎𝑚𝑝𝑙𝑒 𝑚𝑎𝑠𝑠 (𝑔) Eq. (23)

Calculations were performed using Win-Pascal Software Vers. 1.05 with a contact angle of

140° and a surface tension of 0.48 N/m. Low pressure porosimetry was performed with a

Pascal 140 (0 – 400 kPa) and high pressure porosimetry with a Pascal 400 (max. 400 MPa).

Pressure increased after the intruded mercury achieved an equilibration (0.08 – 0.32 kPa/sec).

The dilatometer was first evacuated to 0.01 kPa (≈ 73540 µm) and then filled by mercury.

Afterwards the pressure was increased to 400 MPa (≈ 1.8 nm).

Figure 66 Mercury porosimetry of granule sieve fraction – Example 90 µm

The granules were divided into sieve fraction of 90, 180 and 250 µm by sieve analyses.

Sample size of the granules was about 0.2 g (n = 1). Two different pore size distributions

could be observed over the whole range of radius (pressure range) for all sieve fraction, which

can be attributed to intergranular pores between particles and intragranular pores of particles

(Figure 66) [110,111]. Intergranular pores have been found at a higher radius and

intragranular pores in the region of 0.4 µm – 1.8 µm. Thus cumulative sum of the relative

volume (mm³/g) of the intruded mercury in low radius region was used to compare the

porosity between the granules.

ANALYTICAL METHODS

131

6.2.1.5 Compression analysis

Excipients, blends and granules were tableted by a single punch press (FlexiTab). FlexiTab

provides the possibility to adjust the dwell time, punch velocity and compression pressure

independent of tablet height or mass. Bi-plan punches with a 10 mm diameter were used. Die

filling was done manually with 300 mg of mass for “in-die” compressibility plots (n = 3) and

200 mg for tabletability and compactibility plots. For “in-die” measurements, FlexiTab was

equipped with a force displacement system containing an incremental position sensor and

strain gauges. The force displacement system was corrected to account for an elastic

deformation of the punch with 0.0033 mm/kN and an offset of 0.0148 mm. For details the

reader is referred to Verena Maria Gläßer (2008). Data acquisition was done by DAQ4

(Hucke Software, Solingen, Germany) and data transformation by Tibco Spotfire (Tibco

Software Inc.). The switching threshold from pneumatic to hydraulic pressure was set to

1.2 kN ( 15 MPa). Pressure range for “in-die” measurement was 0 – 235 MPa and for “out-

of-die” 50 – 350 MPa. Filling depth was constant at 8.5 mm. Tablets were weighed directly

after ejection by analytical balance. “Out-of-die” tablets were determined directly after

ejection by an automatic tablet tester to calculate the solid fraction (n = 6) and the tensile

strength (n = 3). Compression settings are depicted in Table 39.

Table 39 Process parameter – FlexiTab – Tableting profile

Upper Punch (downwards) % of

maximal Pressure

Upper punch (upwards) % of maximal

Pressure

20 % Dwell time 50ms 80 %

20 %

90 % 16,2 KN 80 %

15 % 7,13 KN 100 %

Switching threshold from pneumatic to hydraulic pressure 1.2 kN

ANALYTICAL METHODS

132

6.2.1.5.1 Heckel equation determination

Heckel equation is often used to classify materials by compression behaviour [56]. It is

assumed that the densification process follows a first-order law. The course of the

densification process is divided in three phases (Phase1: particle rearrangement,

fragmentation; Phase 2: plastic

deformation; Phase 3: elastic

recovery, decompression) [63].

The Yield pressure is determined

in phase 2 (plastic deformation)

and indicates if a material

undergoes plastic deformation

(low Yield pressure equal to high

compressibility). Calculating the

Yield pressure is done by

applying a linear regression

model at a predefined pressure

range to determine it by 1/slope.

Resulting intercept A of the

linear regression can be seen as

index for particle rearrangement.

Figure 67 Schematic Heckel plot [63]

Heckel = ln(1

1−𝑆𝑜𝑙𝑖𝑑 𝑓𝑟𝑎𝑐𝑡𝑖𝑜𝑛 𝑜𝑟

1

) = 𝑘 ∗ 𝑃 + 𝐴 [61] Eq. (24)

Yield pressure (𝑃𝑦) =1

𝑘 Eq. (25)

For linear regression a defined pressure range was chosen depending on a normal operating

pressure (unprocessed materials 20 MPa – 120 MPa, granules 120 MPa – 160 MPa).

6.2.2 Ribbon – Solid fraction

6.2.2.1 Throughput method

Different authors have described the calculation of the solid fraction of ribbons by weighing

the throughput of the granules [5,26] or the weight of ribbons [27,34]. For comparison all

described formulas were adapted by replacement of ribbon´s weight with granule´s weight

(Peter (2010) [27], Nkansah et al. (2008) [34]).

ANALYTICAL METHODS

133

For this purpose, the complete granule throughput was collected five times for 5 min over

different time intervals after reaching constant gap (steady state).

Herting et al. (2007) [26]:

𝑉𝑜𝑙𝑢𝑚𝑒𝑟𝑖𝑏𝑏𝑜𝑛 = 𝜋 ∗ 𝑑𝑟𝑜𝑙𝑙𝑠 ∗ 𝑤𝑟𝑜𝑙𝑙𝑠 ∗ 𝑣𝑟𝑜𝑙𝑙𝑠 ∗ 𝑔𝑎𝑝 ∗ 𝑡 Eq. (26)

drolls = diameter rolls [cm], wrolls = width rolls [cm], vrolls = velocity rolls [rpm], gap= gap [cm], t = production

time [min]

Nkansah et al. (2008) [34] modified this formula by adding the surface of the rolls and

substituted the gap by height measurements of the ribbons with a micrometer screw:

𝑉𝑜𝑙𝑢𝑚𝑒𝑟𝑖𝑏𝑏𝑜𝑛 = 𝜋 ∗ 𝑑𝑟𝑜𝑙𝑙𝑠 ∗ 𝑤𝑟𝑜𝑙𝑙𝑠 ∗ 𝑣𝑟𝑜𝑙𝑙𝑠 ∗ ℎ𝑒𝑖𝑔ℎ𝑡𝑟𝑖𝑏𝑏𝑜𝑛 ∗ 𝑡 𝑣𝑜𝑖𝑑𝑠𝑢𝑟𝑓𝑎𝑐𝑒 Eq. (27)

voidssurface = knurled Gerteis®: 50 mm width 2.946 cm³, 25 mm width 1.473 cm³

Gamble et al. (2010) modified the formula of Herting for collecting the weight of granules:

𝑉𝑜𝑙𝑢𝑚𝑒𝑟𝑖𝑏𝑏𝑜𝑛 ± (𝑣𝑜𝑖𝑑𝑠𝑢𝑟𝑓𝑎𝑐𝑒 ∗ (𝜋 ∗ 𝑑𝑟𝑜𝑙𝑙𝑠 ∗ 𝑣𝑟𝑜𝑙𝑙𝑠 ∗ 𝑡

𝑔𝑎𝑝)) Eq. (28)

Peter (2010) [27]:

𝑉𝑜𝑙𝑢𝑚𝑒𝑟𝑖𝑏𝑏𝑜𝑛 = (𝜋 ∗ (𝑑𝑟𝑜𝑙𝑙𝑠 + 𝑔𝑎𝑝

2)) ∗ 𝑣𝑟𝑜𝑙𝑙𝑠 ∗ (𝑔𝑎𝑝 ∗ 𝑤𝑟𝑜𝑙𝑙𝑠) Eq. (29)

Solid fraction was calculated according to Eq. (30):

𝑆𝑜𝑙𝑖𝑑 𝑓𝑟𝑎𝑐𝑡𝑖𝑜𝑛𝑟𝑖𝑏𝑏𝑜𝑛=

(𝑚𝑎𝑠𝑠𝑔𝑟𝑎𝑛𝑢𝑙𝑒𝑠 [𝑔]∗ 𝑡

𝑣𝑜𝑙𝑢𝑚𝑒𝑟𝑖𝑏𝑏𝑜𝑛[𝑐𝑚3])

𝑡𝑟𝑢𝑒 𝑑𝑒𝑛𝑠𝑖𝑡𝑦 [𝑔

𝑐𝑚3] Eq. (30)

6.2.2.2 Mercury porosimetry

Win-Pascal Software Vers. 1.05 was applied to calculate the volume of a ribbon with an

contact angle of 140° and a surface tension of 0.48 N/m (see 6.2.1.4). Pressure increased till

an equilibration of the intruded mercury was achieved (0.08 – 0.32 kPa/sec). The dilatometer

was first evacuated to 0.01 kPa (≈ 73540 µm) and then filled by mercury. Afterwards the

pressure increased to 0.150 MPa (pore radius ≈ 4.90µm), as a result only the cracks of the

surface of the ribbons were filled and the mercury surrounded the surface of the ribbon.

Differences between chamber volume and filled dilatometer with mercury and ribbon allowed

to calculate the solid fraction of the ribbon (see Eq. (1)). Ribbons were cut into two pieces of

25 mm width and 10 mm length to fit into the dilatometer (n = 3).

ANALYTICAL METHODS

134

6.2.2.3 GeoPycnometer

Solid fraction measurements were carried out using a GeoPycnometer 1360 which is based on

the principle of volume displacement [17]. The principle, is that the irregularly shaped sample

is surrounded by a graphite powder (Dry Flo) with good flowability. The result is the

displaced volume by the sample ribbon, which allows calculating the solid fraction.

Figure 68 GeoPyc1360® – Micromeritics Instrument Corporation

Ribbons were measured in a rotating cylindrical vessel, whereby a plunger is pushed through

the vessel and a densification takes place until a specified force is gained. The GeoPynometer

calculated the volume increase by a comparison of step 2 and step 3 (see Figure 69).

Figure 69 GeoPycnometry – Process of measurement

ANALYTICAL METHODS

135

Afterwards the solid fraction of the ribbons is calculated by dividing the weight by the

corresponding volume, which represents 𝜌𝐴𝑝𝑎𝑟𝑒𝑛𝑡 𝑑𝑒𝑛𝑠𝑖𝑡𝑦 (see Eq. (1)). Ribbons were cut into

6 rectangular pieces of 25 mm* 30 mm (width / length) in order to fit into the sample cup

(diameter 38.1 mm). Five samples, each containing 6 pieces, for each specific compaction

force, were analysed in a measurement cycle of 6 with a plunger force of 38 N. This method

was used throughout this thesis, except stated otherwise.

6.2.2.4 Near infrared spectroscopy

Infrared spectroscopy can be classified into near infrared, mid infrared and far infrared

spectroscopy. Near infrared (NIR) spectroscopy is a nondestructive spectroscopic analytical

technique. In a wavelength range of 800 nm to 2500 nm, corresponding to a wavenumber of

about 4000 – 12500 cm-1, absorption of the electromagnetic radiation by molecules occurs.

This absorption results in uptake of specific energy, which results to vibrations of chemical

bonds due to overtones and combination vibrations of C-H, O-H and N-H functional groups

[134–136]. Therefore, NIR spectroscopy is also called vibrational spectroscopy. Overtones

vibrations can be described by the anharmonic oscillator model as a transition over multiple

energy levels e.g. from ground stage to a second level occurs. Therefore, these vibrations are

called overtones. In contrast to that, the harmonic oscillator model for fundamental vibrations

is described as transition from one level to the next level [134–136]. Combination vibrations

represent these sums of multiple fundamental vibrations in a NIR spectrum, which are more

intense compared to overtones. The induced overtone and combination vibrations can overlap

resulting in broad peaks in a NIR spectrum. Mathematical techniques (chemometrics) like

multivariate data analysis, principal component analysis (PCA) and partial least square

analysis (PLS) [137–139] are used to investigate chemical information or physical

information. For both properties quantative models can be developed based on a correlation

between concentration of the sample and proportional change of the NIR spectra. Various

authors demonstrated a correlation between NIR spectra to an upcoming physical property

change like particle size distribution of granules [20,38], hardness/porosity of tablets [119–

121] or solid fraction of ribbons [32,38,122,140]. Thus, NIR can be used to investigate the

solid fraction distribution along the roll width between different scales. Process flow of NIR

method development is depicted in Figure 70.

ANALYTICAL METHODS

136

Samples

NIR measurements Reference method

Data pre-processing

Multivariate statistical

method (PCA, PLS)Calibration model

NIR - Method for

unknown samples

Evaluation of

quality by internal

validation:

Correlation of

coefficient

Root mean square

error of cross

validation etc.

Reference values

Figure 70 Process flow – NIR method development

A Fourier-transform NIR spectrometer (Bruker Multi-Purpose Analyzer MPA) equipped with

a fibre optic probe for measurements in diffuse reflection mode was used, which provides a

high precision, accuracy of wavelength measurements and a high scan speed [135,136].

Treatment of acquired spectra was done with OPUS version 7.2.

ANALYTICAL METHODS

137

Diffuse reflection is defined as the ratio of the radiation reflected (I) by the sample compared

to the reflection of a reference surface (I0) (Reflection = I/I0, Absorption = -log (Reflection))

[134,141]. Reflection of the radiation decreases due to fewer boundaries between air and

solids (high solid fraction). Thus, if the solid fraction is high the absorption will increase and

less diffuse reflection is detected [142]. A schematic representation of the NIR radiation and

corresponding reflection of the ribbons, characterised by a different solid fraction, is depicted

in Figure 71.

Figure 71 Schematic drawing - NIR and solid fraction

6.2.2.4.1 Principal component analysis (PCA) and Partial least square analysis (PLS)

Principal component analysis (PCA) allows extracting the relevant information out of the

spectra and gives the opportunity to develop a broader overview of the spectra. Thus, plenty

amount of data can be condensed [139,141]. Observed variables (original data) are converted

to a three-dimensional space [139]. A linear regression is performed to describe the

transformed data, which results into principal components or factors. The main effect, which

explains the highest variance of the data, is defined as the first factor. Higher ranked factors

explain less variance of the data. These factors are described in multi-dimensional factor

space, which leads to a score system of coordinates and enables an interpretation of the

observed variance of the spectra [134]. Partial least square (PLS) analysis allows quantifying

the correlation between observed variance of spectra and the provided properties (reference

ANALYTICAL METHODS

138

values) [134]. Thus, PCA represents an investigation of factors determining the variance

between various spectra, whereby PLS allows correlating the variance of the different spectra

to the observed reference values (e.g., the solid fraction).

6.2.2.4.2 Data transformation

Data pre-processing techniques for NIR spectra are applied for solid samples to minimize

systematic variations caused by interfering light scattering (noisy signal), impact of measuring

setups or sample conditions (see 4.4.1.2.1) [124]. A first derivation was used and calculated

by Savitzky-Golay algorithm by using polynomial fitting over 9 smoothing steps (smoothing

filter) [141].

ANALYTICAL METHODS

139

6.2.3 Tablet

The properties of tablets were measured 48h after production unless stated otherwise.

Geometry and mass of the tablets were obtained by using an automatic tablet tester (Kraemer

UTS 12 F) and an analytic balance ( 1 mg), respectively.

6.2.3.1 Tensile strength

Tensile strength considers the tablet dimensions and is the normalized breaking force

independent of tablet shape. Calculations were done according to USP [16].

𝑇𝑒𝑛𝑠𝑖𝑙𝑒 𝑠𝑡𝑟𝑒𝑛𝑔𝑡ℎ (𝑇𝑆) = 10∗𝐹

𝜋∗𝐷2∗ [

2.84∗ℎ

𝐷−

0.126∗ℎ

𝑊+

3.15∗𝑊

𝐷+ 0.01]

−1

Eq. (31)

F = breaking force [N]; D = diameter [cm]; h = height [cm]; W = central cylinder thickness [cm]

6.2.3.1.1 Reworkability index

An adapted reworkability index was used, which was first introduced by Herting et al. (2007),

to illustrate the loss of tensile strength after roller compaction:

𝑇𝑒𝑛𝑠𝑖𝑙𝑒 𝑠𝑡𝑟𝑒𝑛𝑔𝑡ℎ𝑟𝑎𝑡𝑖𝑜 =𝑇𝑆𝑔𝑟𝑎𝑛𝑢𝑙𝑒

𝑇𝑆𝑝𝑜𝑤𝑑𝑒𝑟(0 𝑘𝑁/𝑐𝑚)[26] Eq. (32)

Adaption for seven compression pressures:

𝑇𝑆𝑟𝑎𝑡𝑖𝑜% =1

𝑛∑ = (𝑛

𝑖=1𝑇𝑆𝑟𝑎𝑡𝑖𝑜 60𝑀𝑃𝑎+𝑇𝑆𝑟𝑎𝑡𝑖𝑜 100 𝑀𝑃𝑎…..+𝑇𝑆𝑟𝑎𝑡𝑖𝑜 300 𝑀𝑃𝑎

𝑛) ∗ 100 Eq. (33)

6.2.3.2 Solid fraction

Solid fraction of tablets was calculated according to Eq. (34):

𝑆𝑜𝑙𝑖𝑑 𝑓𝑟𝑎𝑐𝑡𝑖𝑜𝑛 =𝜌𝐴𝑃𝑃

𝜌𝑇𝑅𝑈𝐸 =

𝑚 [𝑔]

𝑉𝑃[𝑐𝑚3]

𝜌𝑇𝑅𝑈𝐸 [𝑔

𝑐𝑚3] Eq. (34)

Volume round biplane tablet:

𝑉𝑃 = 𝜋 ∗ (𝐷

4)

2

∗ ℎ Eq. (35)

Volume round convex tablet:

𝑉𝑃 = 2 ∗ [1

3∗ 𝜋 ∗ (ℎ𝑐)2 ∗ (3 ∗ 𝑐𝑟) − ℎ𝑐] + (𝜋 ∗ (

𝐷

4)

2

∗ 𝑊) Eq. (36)

𝜌𝐴𝑃𝑃 = apparent density [g/cm³]; 𝜌𝑇𝑅𝑈𝐸 = true density [g/cm³]

VP = volume at applied pressure [cm³]; D = diameter [cm]; h = height [cm]; m = mass [g]; hc = height calotte

[cm]; cr = curvature radius; W = central cylinder thickness [cm]

hc

W

cr

ANALYTICAL METHODS

140

SUMMARY

141

7. SUMMARY

Granulation processes for solid oral dosage forms are commonly used in the pharmaceutical

industry to enhance the quality of the final product, i.e. tablets. Today, roller compaction is

one of the most common granulation techniques for solid oral dosage forms as it provides

advantages like simple operation, due to integrated process control mechanisms, suitability for

water- or heat-sensitive APIs and opportunity for an implementation in a continuous

manufacturing process. Although roller compaction was intensively investigated, the impact

of upscaling from a small to a larger roller compactor or vice versa is not fully understood.

To account for this knowledge gap, in this thesis the effect of a scale up on the quality

attributes of intermediate- and final products, was investigated. Therefore, the controversially

discussed topic of reduced tabletability of roller compacted granules caused by work

hardening phenomena, particle size enlargement effect, porosity of granules and lubricant

sensitivity was investigated. Two formulations, one predominantly plastic deforming and the

other predominantly brittle deforming, were used. Both had been previously characterised in

respect to their compressibility, tabletability and compactibility and processed at both scales

to differentiate between material and scale dependent effects on the intermediate- and final

product. Finally, a successful scale up strategy was developed to achieve the same product

quality for all scales.

Solid fraction of the ribbons is well known as key intermediate critical quality attribute for

downstream processing of a roller compaction process. Different established analytical

methods were compared for the measurement of the solid fraction of ribbons. The

GeoPycnometer method (volume displacement) turned out as the most reliable and most

robust method.

Subsequently, both formulations were dry granulated at both scales with equal process

settings. A higher solid fraction of the ribbons was obtained for both formulations at the larger

scale. For the predominantly plastic deforming formulation the particle size distribution of the

granules was similar for both scales, resulting in a lower tensile strength of the tablets of the

large scale, which was mainly impacted by the work hardening effect and sensitivity towards

lubricant. The increased solid fraction of the ribbon produced by the large scale compared to

the small scale correlated with a lower tensile strength of the tablets. In contrast, negligible

differences of the tensile strength of the tablets between both scales were observed for the

predominantly brittle deforming formulation, although the particle size distribution of the

SUMMARY

142

granules differed at higher specific compaction forces of the large scale. This was driven by

the impact of the brittle deforming component, which enhances the fracturing behaviour of

the granules and resulted in a negligible susceptibility towards work hardening, lubricant and

particle size enlargement effect. In conclusion, even though differences existed between

ribbons produced at both scales, these could be balanced if the formulation contains a high

proportion of a brittle component. This strategy will allow enhancing the robustness of the

scalability of the process and the final product quality.

Previously it was demonstrated that a different solid fraction of the ribbon resulted in a

different tensile strength of the tablets between scales for the predominantly plastic

formulation. This formulation however, is commonly used to counteract the main

disadvantage of the roller compaction; the reduced tabletability of granules of tablets (loss of

tensile strength). To account for this, a new approach (Scale Model) was developed for the

predominantly plastic formulation to achieve the same solid fraction of the ribbon at both

scales. Same solid fraction at both scales resulted in a same porosity of the granules,

compressibility and tensile strength of the tablets, although a different particle size

distribution of the granules was obtained. This demonstrated that the particle size distribution

of granules should not to be considered as the main intermediate quality attribute to achieve a

successful scale up for a roller compaction process, because the porosity and the

compressibility of the granules defining the microstructure of a tablet during tableting and

subsequently the resulting tensile strength of the tablets. The Scale Model approach

demonstrated a practicable solution for the pharmaceutical industry to scale the process from

small development batches to commercial batches and still achieve equal quality of the

tablets.

In order to investigate the observed higher solid fraction at the large scale at equal process

settings for both scales a new analytical method via NIR was developed to measure the solid

fraction distribution along the roll width. It was possible to predict the solid fraction of

unknown samples by acquiring the NIR spectra comprising reduction of analysis time

compared to the GeoPycnometer method, which measured the “total” solid fraction of the

ribbon. The effect of the cheek plates (lower solid fraction at the edges) decreased with

increased distance to the cheek plates, which was especially the case for the larger scale with

a broader roll width. This led to a higher “total” solid fraction of the ribbons produced by the

large scale compared to the small scale at equal process settings. These results explained the

previously observed different quality attributes of intermediate- and final products. The

proposed scale up approach showed that the differences of resulting granules and tablets

SUMMARY

143

between scales can be balanced for a predominantly plastic deforming formulation through

the adaption of the specific compaction force. Thus, adapting the specific compaction force by

measurements of the “total” solid fraction (GeoPycnometer) is a suitable scale up strategy for

a roller compaction process.

Moreover, the solid fraction of a tablet (compressibility) was an important impact factor,

which reinforced the development of theoretical models to predict the solid fraction for

unknown powder mixtures based on single component compression analysis. A new

theoretical developed Percolation and a modified Kawakita model were evaluated for model

application. An exponential model was added to elucidate whether the two-parametrised

models with theoretical background are superior in terms of predictability of solid fraction

compared to a model without parametrised variables. Four mixtures were compressed over a

wide pressure range at various fractions of a plastic and brittle deforming component. Based

on single compression analysis of the pure excipients and application of these models, it was

possible to predict the solid fraction of all mixtures. The Kawakita model showed overall

superior prediction accuracy, whereas the Percolation model resulted in the best fit for

mixtures containing the plastic deforming component in a range of 72%–48%. Both models

were in good agreement at residuals below 3%. The prediction could serve as a systematic

guidance for the formulator to select appropriate excipients depending on the active

pharmaceutical ingredient to build quality into the drug product according to the Quality by

Design approach.

In summary, this thesis provides a new profound knowledge and an appropriate guidance for

the scale up of a roller compaction process. An effect of a scale up of a roller compaction

process on the quality attributes of intermediate- and final products was demonstrated. This

effect can be balanced by applying the proposed scale up strategy or by diminishing the

formulation susceptibility to scale dependent effects with an increased proportion of a

predominantly brittle deforming component in the formulation.

SUMMARY

144

APPENDIX

145

8. APPENDIX

8.1 ANALYTIC DATA


COMPARISON FOR SOLID FRACTION MEASUREMENTS OF

RIBBONS


Mean height ribbon (Nkansah et al. (2008), micrometer screw):

Specific compaction force

[kN/cm] Gap [mm]

Mean height ribbon [mm]

SD

2 3 3.585 0.027

4 3 3.798 0.047

6 3 3.790 0.047

8 3 3.806 0.022

SD = Standard deviation of mean

APPENDIX

146

8.1.2 CHAPTER 4.2 COMPARISON OF TWO ROLLER COMPACTORS OF

DIFFERENT SCALE AT SAME PROCESS


Granule attributes (MacroPactor):

Specific

compaction

force

[kN/cm]


d50 [µm]

SD

n = 3

Hausner

ratio*US

P n = 1

Bulk

density

[g/cm³]

n = 1

d50 [µm]

SD

n = 3

Hausner

ratio*US

P n = 1

Bulk

density

[g/cm³]

n = 1

0 103.00

0.00 1.23 0.48

110.00

2.05 1.23 0.52

2 83.17

1.41 1.29 0.51

78.76

3.56 1.27 0.53

4 97.80

0.82 1.35 0.53

81.51

2.95 1.33 0.56

6 138.13

13.36 1.32 0.54

139.13

2.94 1.31 0.58

8 132.75

11.95 1.26 0.56

210.17

2.45 1.31 0.59

MCC = Microcrystalline cellulose; LAC = − Lactosemonohydrate; SD = Standard deviation of mean

APPENDIX

147

Granule attributes (MiniPactor):

Specific

compaction

force

[kN/cm]


d50 [µm]

SD

n = 3

Hausner

ratio*US

P n = 1

Bulk

density

[g/cm³]

n = 1

d50 [µm]

SD

n = 3

Hausner

ratio*US

P n = 1

Bulk

density

[g/cm³]

0 103.00

0.00 1.23 0.48

110.00

2.05 1.23 0.52

2 87.56

0.47 1.27 0.47

70.35

0.47 1.37 0.50

4 98.25

0.00 1.33 0.50

73.04

0.82 1.37 0.55

6 112.60

1.24 1.31 0.54

88.80

3.68 1.35 0.59

8 136.38

2.45 1.31 0.56

105.35

0.00 1.33 0.60

MCC = Microcrystalline cellulose; LAC = − Lactosemonohydrate; SD = Standard deviation of mean

Tablets comparison milling process at two scale: Solid fraction tablets

Formulation SF MacroPactor

SF SD n = 20

MiniPactor

SF SD n = 20

MCC 2:1 LAC

0.62 0.6239 0.0032 0.6253 0.0042

0.77 0.7748 0.0036 0.7759 0.0028

MCC 1:1 LAC

0.64 0.6370 0.0033 0.6367 0.0037

0.79 0.7921 0.0016 0.7889 0.0030

MCC = Microcrystalline cellulose; LAC = − Lactosemonohydrate; SF = Solid fraction; SD = Standard

deviation of mean

APPENDIX

148

4.2.4 Influence of granules on tablet attributes

Reworkability index tablets – MiniPactor – MCC 2:1 LAC/ MCC 1:1 LAC

Reworkability index tablets –MiniPactor – MCC 2:1 LAC/MCC 1:1 LAC, TSratio = tensile strength of tablets of

compacted blend in proportion to unprocessed blend Eq. (33), mean (n = 350), error bars (standard deviation of

mean)

Results Heckel -Yield pressure granules – MacroPactor/MiniPactor – MCC 1:1 LAC

Yield

pressure

Py (1/slope)

0 kN/cm 2 kN/cm 4 kN/cm 6 kN/cm 8 kN/cm

MacroPactor 189.68 197.90 201.12 208.36 211.87

MiniPactor 189.68 197.18 200.73 206.72 212.57

Difference 0 0.72 0.39 1.64 -0.70

Py = Yield pressure Heckel

APPENDIX

149

8.1.3 CHAPTER 4.4 INVESTIGATION OF SOLID FRACTION DISTRIBUTION

ALONG THE ROLL WIDTH BETWEEN DIFFERENT SCALES VIA NIR

AT-LINE

4.4.1 Method evaluation to determine solid fraction along the roll width by GeoPycnometer

and NIR

Spectral bands NIR spectra:

Content Chemical group Specified range

Metformin[125]

Primary amine NH2 - Combination

bands

5000 – 4762 cm-1, (2000 –

2100 nm)

Secondary amine NH - Overtone

bands

6570 cm-1

(1522 nm),

9597 cm-1,

(1042 nm)

Imid group C=NH –

Overtone band

5924 cm-1,

(1688 nm)

Water[126]

Hydroxyl O–H –

Combination bands

5208 cm-1,

(1920 nm)

Hydroxyl O–H –

Overtone band

7040 cm-1

(1420 nm)

APPENDIX

150

8.1.4 MODEL DEVELOPMENT – PREDICTING SOLID FRACTION OF A

TABLET

4.5.1 Application of models – Excipients

Correlation Kawakita C and new derived C solid fraction

Material V0 [cm³] at 1-2

MPa

Correlation

Kawakita C vs.

Derived C solid

fraction

R2

MCC 0.398955 0.4647 x + 0.2255 0.9951

LAC 0.211695 0.9113 x + 0.4904 0.9954

CARB 0.267273 0.7789 x + 0.2133 0.9928


V0 = Initial volume [cm³]; C = Degree of volume reduction; R2 = Correlation coefficient

APPENDIX

151

8.2 STATISTICAL TESTS


COMPARISON FOR SOLID FRACTION MEASUREMENTS OF

RIBBONS

4.1.2 Comparison of throughput, mercury porosimetry and GeoPycnometry

Comparison mercury porosimetry and GeoPycnometry – T-Test

Specific compaction force [kN/cm]

Mercury porosimetry (n = 3) vs.

Geopycnometry (n = 5)

T- Test α = 0.05

p

2 0.1471

4 0.8437

6 0.7130

8 0.1614

APPENDIX

152

8.2.2 CHAPTER 4.2 COMPARISON OF TWO ROLLER COMPACTORS OF

DIFFERENT SCALE AT SAME PROCESS


Solid fraction comparison between formulations within one scale – T-Test

Specific compaction force

[kN/cm]

MCC 2:1 LAC vs. MCC

1:1 LAC

MacroPactor

T- Test α = 0.05

p

MCC 2:1 LAC vs. MCC

1:1 LAC

MiniPactor

T- Test α = 0.05

p

2 0.000000 0.000075

4 0.000000 0.000000

6 0.000001 0.000000

8 0.000001 0.000034


APPENDIX

153


d50 granules – MacroPactor/MiniPactor- MCC 2:1 LAC/MCC 1:1 LAC- T-Test

Formulation

Specific

compaction

force [kN/cm]

d50 [µm]

SD n = 3 T-test α = 0.05

p MacroPactor MiniPactor

MCC 2:1 LAC

0 103.00 0.00 -

2 83.17 1.41 87.56 0.47 0.0147

4 97.80 0.82 98.25 0.00 1.0000

6 138.13 13.36 112.60 1.25 0.0501

8 132.75 11.95 136.38 2.45 0.7189

MCC 1:1 LAC

0 110.00 2.05 -

2 78.76 3.56 70.35 0.47 0.0269

4 81.51 2.95 73.04 0.82 0.0111

6 139.13 2.94 88.80 3.68 0.0001

8 210.17 2.45 105.35 0.00 0.0000

MCC = Microcrystalline cellulose; LAC = − Lactosemonohydrate; d50 = Median particle dimension; SD =

Standard deviation of mean

d50 granules of milled tablets – MacroPactor/MiniPactor- MCC 2:1 LAC/MCC 1:1 LAC- T-

Test

Formulation Milled tablets

(SF)

MacroPactor

d50 [µm] SD

n = 3

MiniPactor

d50 [µm] SD n

= 3

T-Test

α = 0.05

p

MCC 2:1 LAC

0.62 99.16 4.96 99.55 1.89 0.9336

0.77 163.16 7.41 163.98 10.27 0.9721

MCC 1:1 LAC

0.64 86.41 0.47 88.45 1.70 0.1347

0.79 138.76 6.94 145.80 3.40 0.2696

MCC = Microcrystalline cellulose; LAC = − Lactosemonohydrate; d50 = Median particle dimension; SD =

Standard deviation of mean

APPENDIX

154

4.2.4 Influence of granules on tablet

Pre-validation sample size tablets (tensile strength)

A Pre-validation of the tensile strength showed a relative standard deviation of maximum

10 % within one compression pressure. Thus, differences of over 10 % of tensile strength

were defined as relevant for scale up. Target was to detect these differences with a power of

0.95 over the whole compression pressure range. Sample size 50 tablets for each compression

pressure. (Pre-validation setup: MCC 2:1 LAC, 4 kN/cm)

Compression

pressure [MPa] 59.68 99.47 139.26 179.05 218.84 258.63 298.42

Tensile strength

[N/mm²] 0.39 0.94 1.53 2.03 2.47 2.88 3.02

Standard deviation

of mean 0.0392 0.087 0.1389 0.1448 0.1482 0.1752 0.2033

Relative standard-

deviation % 10 9.3 9.09 7.14 6 6.09 6.72

Detect differences if

value =

(+10 %)

0.43 1.03 1.68 2.23 2.72 3.17 3.33

Detect differences if

value =

(-10 %)

0.35 0.84 1.37 1.82 2.22 2.59 2.72

Calculated sample

size to detect

difference 10 %

with a Power of 0.95

(1-beta) alpha = 0.05

52 52 48 30 22 22 26

Determined sample

size of 50 resulting

Power for 10 %

deviation

0.94 0.95 0.96 1 1 1 1

APPENDIX

155

One Way ANOVA within scale and compression pressure (equal to five means comparison

with each other)

All tests were passed (see column comment)

A. Levenes’s test for homogeneity (α = 0.05)

B. ANOVA univariate test of significance (α = 0.05)

C. Post Hoc ANOVA Bonferroni (α = 0.05)

MacroPactor Compression

Pressure Comment

0

kN/cm

2

kN/cm

4

kN/cm

6

kN/cm

8

kN/cm

MCC 2:1 LAC

59.68 * * X X *

99.47 * * * * *

139.26 * * * * *

179.05 * * * * *

218.84 * * * X X

258.63 * * * * *

298.42 * * X X *

MCC 1:1 LAC

59.68 * * X * X

99.47 * * * * *

139.26 * * * * *

179.05 * * * * *

218.84 + * * * X X

258.63 + * * * * *

298.42 * * * X X


* indicates significanct differences (p 0.05)

X = no significant difference between values

+ Levene’s test showed significance (eq. heteroscedasticity)

APPENDIX

156

MiniPactor Compression

Pressure Comment

0

kN/cm

2

kN/cm

4

kN/cm

6

kN/cm

8

kN/cm

MCC 2:1 LAC

59.68 * * * * *

99.47 + * * * * *

139.26 * * * * *

179.05 * * * * *

218.84 * * * * *

258.63 * * * * *

298.42 * * * * *

MCC 1:1 LAC

59.68 * * * X X

99.47 * * * * *

139.26 * * * * *

179.05 + * * * * *

218.84 + * * * * *

258.63 + * * * * *

298.42 * * * * *


* indicates significant differences (p 0.05)


+ Levene’s test showed significance (eq. heteroscedasticity)

APPENDIX

157

Tensile strength tablets - Impact scale - MacroPactor vs. MiniPactor – MCC 2:1 LAC/MCC

1:1 LAC - T-Test

Formulation

Compression

pressure

[MPa]

Tensile strength [N/mm²] n = 50

T-test α = 0.05

MacroPactor vs. MiniPactor

Comment 2 kN/cm 4 kN/cm 6 kN/cm 8 kN/cm

MCC 2:1 LAC

59.68 * * X X

99.47 * * * * <10 %

139.26 * * * *

179.05 * * * *

218.84 * * * * <10 %

258.63 * * * <10 % *

298.42 * * *+ * <10 %

MCC 1:1 LAC

59.68 * <10 % X <10 % * <10 % +* <10 %

99.47 * <10 % * X <10 % X <10 %

139.26 +* * <10 % X <10 % * <10 %

179.05 * <10 % * <10 % * <10 % +* <10 %

218.84 X <10 % * <10 % X <10 % * <10 %

258.63 X <10 % * <10 % X <10 % +* <10 %

298.42 X <10 % X <10 % X <10 % X <10 %


* = indicates significant differences (p 0.05)


+ = Levene’s test showed significance (eq. heteroscedasticity)

<10 % = mean values difference smaller than 10 %

APPENDIX

158

8.2.3 CHAPTER 4.3 ADAPTED PROCESS SETTINGS OF DIFFERENT

SCALES TO ACHIEVE SIMILAR PRODUCT QUALITY

4.3.1 Achieving the same solid fraction of ribbon by using adapted process parameter

Solid fraction ribbons – MacroPactor/Scale Model – MCC 2:1 LAC – T-Test


Target SF vs. MCC 2:1 LAC Scale Model

T- Test α = 0.05

p

2 0.000000

4 0.636793

6 0.000000

8 0.000000

SF = Solid fraction; MCC = Microcrystalline cellulose; LAC = − Lactosemonohydrate

4.3.2 Particle size distribution and porosity of granules

d50 granules – MacroPactor/Scale Model- MCC 2:1 LAC - T-Test

MacroPactor Scale Model

T-Test α = 0.05

p

Specific

compaction

force kN/cm

d50 [µm] SD

n = 3

Specific

compaction

force [kN/cm]

d50 [µm] SD

n = 3

2 83 1 4.1 105 0 0.000033

4 98 1 7.0 137 1 0.000005

6 138 13 9.6 179 2 0.013335

8 133 12 11.4 245 6 0.000294

d50 = Median particle dimension; SD = Standard deviation of mean

APPENDIX

159

4.3.3 Tabletability and compressibility influenced by material attributes of granules and

ribbons

Example calculation of detecting a 5 % difference with a sample size of n = 50

Compression

pressure [MPa] 59.68 99.47 139.26 179.05 218.84 258.63 298.42

Tensile strength

[N/mm²] 0.39 0.94 1.53 2.03 2.47 2.88 3.02

Standard

deviation of mean 0.0392 0.087 0.1389 0.1448 0.1482 0.1752 0.2033

Relative standard

deviation % 10 9.3 9.09 7.14 6 6.09 6.72

Detect differences

if value =

(+5 %)

0.41 0.98 1.6 2.13 2.59 3.02 3.18

Detect differences

if value =

(-5 %)

0.37 0.89 1.45 1.93 2.35 2.73 2.87

Determined

sample size of 50

resulting Power

for ±5 % deviation

0.21 0.62 0.7 0.93 0.98 0.93 0.97

APPENDIX

160

Tensile strength tablets - MacroPactor vs. Scale Model – MCC 2:1 LAC/MCC 1:1 LAC - T-

Test

Formulation

Compression

pressure

[MPa]

Tensile strength [N/mm²] n = 50

T-test α = 0.05

MacroPactor vs. Scale Model

Comment 2 kN/cm 4 kN/cm 6 kN/cm 8 kN/cm

MCC 2:1 LAC

59.68 * *<10 % * *

99.47 *<10 % *<10 % * *

139.26 *<10 % *<10 % * *

179.05 + X<10 % *<10 % * *

218.84 X<10 % *<10 % X<10 % *

258.63 X<10 % *<10 % * *

298.42 *<10 % *<10 % *<10 % *


* = indicates significant differences (p 0.05)


+ = Levene’s test showed significance (eq. heteroscedasticity)

<10 % = mean values difference smaller than 10 %

APPENDIX

161

8.2.4 CHAPTER 4.4 INVESTIGATION OF SOLID FRACTION DISTRIBUTION

ALONG THE ROLL WIDTH BETWEEN DIFFERENT SCALES VIA NIR

4.4.3 Scale up approach – Comparison of solid fractions of ribbons of a Metformin

formulation at two

Solid fraction ribbons – MacroPactor/MiniPactor – T-Test


MacroPactor MET 21

vs.

MiniPactor MET 21

T- Test α = 0.05

p

2 0.000014

4 0.001711

6 0.000041

8 0.000046

MET 21 = Metformin drug load 21 %

APPENDIX

162

8.3 LIST OF TABLES

Table 1 Overview objectives ........................................................................................... 3

Table 2 Material attributes - Raw materials and blends ................................................ 19

Table 3 Results Heckel – Excipients and blends ........................................................... 22

Table 4 Summary - Compression analysis excipients and blends – Descending order

of maximal achievable tabletability and compactibility ................................... 25

Table 5 d50 milled tablets – MacroPactor/MiniPactor – MCC 2:1 LAC/MCC 1:1

LAC .................................................................................................................. 43

Table 6 Heckel - Yield pressure granules - MiniPactor/MacroPactor - MCC 2:1

LAC .................................................................................................................. 51

Table 7 Summary - Evaluation NIR setup ..................................................................... 87

Table 8 Summary – Results internal validation ............................................................. 91

Table 9 Summary - Results external validation ............................................................ 93

Table 10 Calculated true density mixtures .................................................................... 103

Table 11 Estimated parameters for the Percolation model ............................................ 104

Table 12 Estimated parameters for the Kawakita model ............................................... 107

Table 13 Estimated parameters for the Exponential model ........................................... 107

Table 14 Materials ......................................................................................................... 111

Table 15 Placebo composition – Chapter 4.1, 4.2 and 4.3 ............................................ 113

Table 16 API composition (Metformin) – Chapter 4.4 ................................................. 113

Table 17 Placebo composition – Chapter 4.5 ................................................................ 114

Table 18 Overview experimental methods .................................................................... 115

Table 19 Manufacturing equipment ............................................................................... 116

Table 20 Manufacturing flow chart: Mixture ................................................................ 117

Table 21 Process parameters: Mixture........................................................................... 118

Table 22 Batch size: Mixture ......................................................................................... 118

Table 23 Manufacturing flow chart: Chapter 4.2 and 4.3 .............................................. 119

Table 24 Process parameters: Chapter 4.2 and 4.3 ........................................................ 120

Table 25 Batch size: Chapter 4.2 and 4.3 ...................................................................... 121

Table 26 Manufacturing flow chart: Chapter 4.2: Comparison milling process ........... 121

Table 27 Process parameters: Chapter 4.2: Comparison milling process ..................... 122

Table 28 Batch size: Comparison milling process......................................................... 122

APPENDIX

163

Table 29 Manufacturing flow chart: Chapter 4.3: Porosity granules ............................ 123

Table 30 Process parameters: Chapter 4.3: Porosity granules ....................................... 123

Table 31 Manufacturing flow chart: Chapter 4.4 .......................................................... 124

Table 32 Process Parameters Chapter 4.4 ...................................................................... 124

Table 33 Batch size: Chapter 4.4 ................................................................................... 124

Table 34 Manufacturing flow chart: Chapter 4.5 .......................................................... 125

Table 35 Process parameters: Chapter 4.5 ..................................................................... 126

Table 36 Batch size: Chapter 4.5 ................................................................................... 126

Table 37 Analytical equipment ...................................................................................... 127

Table 38 Software and data processing ......................................................................... 128

Table 39 Process parameter – FlexiTab – Tableting profile .......................................... 131

REFERENCES

164

REFERENCES

[1] Dilip M. Parikh (Ed.), Handbook of Pharmaceutical Granulation Technology, Second

Edition ed., Taylor & Friends, 2005.

[2] P. Kleinebudde, Roll compaction/dry granulation: pharmaceutical applications,

European Journal of Pharmaceutics and Biopharmaceutics 58 (2004) 317–326.

[3] Y. Teng, Z. Qiu, H. Wen, Systematical approach of formulation and process

development using roller compaction, European Journal of Pharmaceutics and

Biopharmaceutics 73 (2009) 219–229.

[4] R.M. Iyer, S. Hegde, J. DiNunzio, D. Singhal, W. Malick, The impact of roller

compaction and tablet compression on physicomechanical properties of

pharmaceutical excipients, Pharmaceutical Development and Technology 19 (2014)

583–592.

[5] J.F. Gamble, M. Tobyn, A.B. Dennis, T. Shah, Roller compaction: application of an

in-gap ribbon porosity calculation for the optimization of downstream granule flow

and compactability characteristics, Pharmaceutical Development and Technology 15

(2010) 223–229.

[6] B. Rambali, L. Baert, E. Jans, D. Massart, Influence of the roll compactor parameter

settings and the compression pressure on the buccal bio-adhesive tablet properties,

International Journal of Pharmaceutics 220 (2001) 129–140.

[7] S.K. Kochhar, M.H. Rubinstein, D. Barnes, The effects of slugging and recompression

on pharmaceutical excipients, International Journal of Pharmaceutics 115 (1995) 35–

43.

[8] Hancock, B.C., Colvin, J.T., Mullarney, M.P., Zinchuk, A.V., The relative densities of

pharmaceutical powders, blends, dry granulations, and immediate-release tablets,

Pharmaceutical Technology (2003) 1543–2521.

[9] J. Bultmann, Multiple compaction of microcrystalline cellulose in a roller compactor,


[10] J. Mosig, P. Kleinebudde, Critical evaluation of root causes of the reduced

compactability after roll compaction/dry granulation, Journal of Pharmaceutical

Sciences 104 (2015) 1108–1118.

[11] G. Bindhumadhavan, J. Seville, M.J. Adams, R.W. Greenwood, S. Fitzpatrick, Roll

compaction of a pharmaceutical excipient: Experimental validation of rolling theory

for granular solids, Chemical Engineering Science 60 (2005) 3891–3897.

REFERENCES

165

[12] M. Bi, F. Alvarez-Nunez, F. Alvarez, Evaluating and modifying Johanson's rolling

model to improve its predictability, Journal of Pharmaceutical Sciences 103 (2014)

2062–2071.

[13] Arthi D. Rajkumar, Gavin K. Reynolds, David Wilson, Stephen Wren, Michael J.

Hounslow, Agba D. Salman, Investigating the effect of processing parameters on

pharmaceutical tablet disintegration using a real-time particle imaging approach,


[14] Julian Quodbach & Peter Kleinebudde, A critical review on tablet disintegration,

Pharmaceutical Development and Technology 21 (2016) 763–774.

[15] L.X. Yu, G. Amidon, M.A. Khan, S.W. Hoag, J. Polli, G.K. Raju, J. Woodcock,

Understanding pharmaceutical quality by design, The AAPS journal 16 (2014) 771–

783.

[16] United States Pharmacopeial Convention Inc., United States Pharmacopeia - National

Formulary, USP 39–NF 34 ed.

[17] A.V. Zinchuk, M.P. Mullarney, B.C. Hancock, Simulation of roller compaction using

a laboratory scale compaction simulator, International Journal of Pharmaceutics 269

(2004) 403–415.

[18] ICH, Q8: Pharmaceutical development: Part I: pharmaceuticaldevelopment; Part II:

annex to pharmaceutical development., 2009,

http://www.ich.org/fileadmin/Public_Web_Site/ICH_Products/Guidelines/Quality/Q8_

R1/Step4/Q8_R2_Guideline.pdf, accessed 4 June 2017.

[19] C.S. Omar, R.M. Dhenge, J.D. Osborne, T.O. Althaus, S. Palzer, M.J. Hounslow, A.D.

Salman, Roller compaction: Effect of morphology and amorphous content of lactose

powder on product quality, International Journal of Pharmaceutics 496 (2015) 63–74.

[20] R. Kona, R.M. Fahmy, G. Claycamp, J.E. Polli, M. Martinez, S.W. Hoag, Quality-by-

design III: application of near-infrared spectroscopy to monitor roller compaction in-

process and product quality attributes of immediate release tablets, AAPS

PharmSciTech 16 (2015) 202–216.

[21] H. Lim, V.S. Dave, L. Kidder, E. Neil Lewis, R. Fahmy, S.W. Hoag, Assessment of

the critical factors affecting the porosity of roller compacted ribbons and the feasibility

of using NIR chemical imaging to evaluate the porosity distribution, International

Journal of Pharmaceutics 410 (2011) 1–8.

[22] M. Khorasani, J.M. Amigo, J. Sonnergaard, P. Olsen, P. Bertelsen, J. Rantanen,

Visualization and prediction of porosity in roller compacted ribbons with near-infrared

REFERENCES

166

chemical imaging (NIR-CI), Journal of Pharmaceutical and Biomedical Analysis 109

(2015) 11–17.

[23] M.K. Ghorab, R. Chatlapalli, S. Hasan, A. Nagi, Application of thermal effusivity as a

process analytical technology tool for monitoring and control of the roller compaction

process, AAPS PharmSciTech 8 (2007) 23.

[24] J. Sakwanichol, S. Puttipipatkhachorn, G. Ingenerf, P. Kleinebudde, Roll

compaction/dry granulation: comparison between roll mill and oscillating granulator in

dry granulation, Pharmaceutical Development and Technology 17 (2012) 30–39.

[25] J. Nordstrom, G. Alderborn, The granule porosity controls the loss of compactibility

for both dry- and wet-processed cellulose granules but at different rate, Journal of

Pharmaceutical Sciences 104 (2015) 2029–2039.

[26] M.G. Herting, P. Kleinebudde, Roll compaction/dry granulation: effect of raw material

particle size on granule and tablet properties, International Journal of Pharmaceutics

338 (2007) 110–118.

[27] S. Peter, Untersuchungen zur Homogenisierung der Schülpendichte und Entwicklung

eines Vorhersagemodells auf der Basis von Tablettierversuchen. PhD thesis,

University of Bonn, 2010.

[28] R.F. Mansa, Roll compaction of pharmaceutical excipients and prediction using

intelligent software. PhD thesis, University of Birmingham, 2006.

[29] T. Feng, F. Wang, R. Pinal, C. Wassgren, M.T. Carvajal, Investigation of the

variability of NIR in-line monitoring of roller compaction process by using Fast

Fourier Transform (FFT) analysis, AAPS PharmSciTech 9 (2008) 419–424.

[30] A.M. Miguelez-Moran, C.-Y. Wu, H. Dong, J.P.K. Seville, Characterisation of density

distributions in roller-compacted ribbons using micro-indentation and X-ray micro-

computed tomography, European Journal of Pharmaceutics and Biopharmaceutics 72

(2009) 173–182.

[31] I. Akseli, S. Iyer, H.P. Lee, A.M. Cuitino, A quantitative correlation of the effect of

density distributions in roller-compacted ribbons on the mechanical properties of

tablets using ultrasonics and X-ray tomography, AAPS PharmSciTech 12 (2011) 834–

853.

[32] D. Acevedo, A. Muliadi, A. Giridhar, J.D. Litster, R.J. Romanach, Evaluation of three

approaches for real-time monitoring of roller compaction with near-infrared

spectroscopy, AAPS PharmSciTech 13 (2012) 1005–1012.

REFERENCES

167

[33] F. Wöll, Entwicklung von Methoden zur Charakterisierung von Schülpen. PhD thesis,

University of Halle-Wittenberg, 2003.

[34] P. Nkansah, S.-J. Wu, S. Sobotka, K. Yamamoto, Z.J. Shao, A novel method for

estimating solid fraction of roller-compacted ribbons, Drug Development and

Industrial Pharmacy 34 (2008) 142–148.

[35] J. Brudy, Systematische Rezepturentwicklung für die Walzenkompaktierung: Anhand

binärer und ternärer Mischungen ausgewählter Komponenten. PhD thesis, University

of Heidelberg, 2007.

[36] A. Michrafy, H. Diarra, J.A. Dodds, M. Michrafy, Experimental and numerical

analyses of homogeneity over strip width in roll compaction, Powder Technology 206

(2011) 154–160.

[37] M. Alleso, R. Holm, P. Holm, Roller compaction scale-up using roll width as scale

factor and laser-based determined ribbon porosity as critical material attribute,

European journal of pharmaceutical sciences official journal of the European

Federation for Pharmaceutical Sciences 87 (2016) 69–78.

[38] Abhay Gupta, Garnet E. Peck, Ronald W. Miller, Kenneth R. Morris, Nondestructive

measurements of the compact strength and the particle-size distribution after milling of

roller compacted powders by near-infrared spectroscopy, Journal of Pharmaceutical

Sciences 93 (2004) 1047–1053.

[39] P. Guigon, O. Simon, Roll press design—influence of force feed systems on

compaction, Powder Technology 130 (2003) 41–48.

[40] Y. Funakoshi, T. Asogawa, E. Satake, The use of a Novel Roller Compactor with a

Concavo-Convex Roller pair to Obtain Uniform Compacting Pressure, Drug

Development and Industrial Pharmacy 3 (2008) 555–573.

[41] A.M. Miguelez-Moran, C.-Y. Wu, J.P.K. Seville, The effect of lubrication on density

distributions of roller compacted ribbons, International Journal of Pharmaceutics 362

(2008) 52–59.

[42] W. Shi, O.L. Sprockel, A practical approach for the scale-up of roller compaction

process, European Journal of Pharmaceutics and Biopharmaceutics 106 (2016) 15–19.

[43] Ana Pérez Gago, Gavin Reynolds, Peter Kleinebudde, Impact of roll compactor scale

on ribbon density. In Press, Corrected Proof, Powder Technology (2017).

[44] S. Malkowska, K.A. Khan, Effect of Re-Conpression on the Properties of Tablets

Prepared by Dry Granulation, Drug Development and Industrial Pharmacy 9 (1983)

331–347.

REFERENCES

168

[45] Christer Nyström, Göran Alderborn, Margareta Duberg & Per-Gunnar Karehill,

Bonding Surface area and Bonding Mechanism -Two Important Factors fir the

Understanding of Powder Comparability, Drug Development and Industrial Pharmacy

19 (1993) 2143–2196.

[46] C.C. Sun, M.W. Himmelspach, Reduced tabletability of roller compacted granules as a

result of granule size enlargement, Journal of Pharmaceutical Sciences 95 (2006) 200–

206.

[47] M.G. Herting, P. Kleinebudde, Studies on the reduction of tensile strength of tablets

after roll compaction/dry granulation, European Journal of Pharmaceutics and


[48] S.-J. Wu, C.C. Sun, Insensitivity of compaction properties of brittle granules to size

enlargement by roller compaction, Journal of Pharmaceutical Sciences 96 (2007)

1445–1450.

[49] X. He, P.J. Secreast, G.E. Amidon, Mechanistic study of the effect of roller

compaction and lubricant on tablet mechanical strength, Journal of Pharmaceutical

Sciences 96 (2007) 1342–1355.

[50] C.C. Sun, P. Kleinebudde, Mini review: Mechanisms to the loss of tabletability by dry

granulation, European Journal of Pharmaceutics and Biopharmaceutics 106 (2016) 9–

14.

[51] Narang, Ajit S, Boddu, Sai H S. (Ed.), Excipient Applications in Formulation Design

and Drug Delivery, Springer International Publishing AG Switzerland, 2015.

[52] Saurabh M. Mishra & Bhagwan D. Rohera, An integrated, quality by design (QbD)

approach for design, development and optimization of orally disintegrating tablet

formulation of carbamazepine, Pharmaceutical Development and Technology 22

(2017) 889–903.

[53] C.K. Tye, C.C. Sun, G.E. Amidon, Evaluation of the effects of tableting speed on the

relationships between compaction pressure, tablet tensile strength, and tablet solid

fraction, Journal of Pharmaceutical Sciences 94 (2005) 465–472.

[54] J.M. Sonnergaard, A critical evaluation of the Heckel equation (1999) 63–71.

[55] C. Sun, D.J. Grant, Influence of Elastic Deformation of Particles on Heckel Analysis,


[56] Jukka Ilkka and Petteri Paronen, Prediction of the compression behaviour of powder

mixtures by the Heckel equation, International Journal of Pharmaceutics 94 (1993)

181–187.

REFERENCES

169

[57] Fredrik Nicklasson, Göran Alderborn, Analysis of the Compression Mechanics of

Pharmaceutical Agglomerates of Different Porosity and Composition Using the Adams

and Kawakita Equations, Pharmaceutical Research 17 (2000).

[58] G. Frenning, J. Nordström, G. Alderborn, Effective Kawakita parameters for binary

mixtures, Powder Technology 189 (2009) 270–275.

[59] J. Nordström, K. Welch, G. Frenning, G. Alderborn, On the physical interpretation of

the Kawakita and Adams parameters derived from confined compression of granular

solids, Powder Technology 182 (2008) 424–435.

[60] S. Patel, A.M. Kaushal, A.K. Bansal, Compression Physics in the Formulation

Development of Tablets, Crit Rev Ther Drug Carrier Syst 23 (2006) 1–66.

[61] Heckel RW., Density–pressure relationship in powder compaction, Transactions of the

Metallurgical Society Of Aime (1961) 671–675.

[62] K. Kawakita, K.-H. Lüdde, Some considerations on powder compression equations,

Powder Technology 4 (1971) 61–68.

[63] M. Duberg, C. Nyström, Studies on direct compression of tablets XVII. Porosity—

pressure curves for the characterization of volume reduction mechanisms in powder

compression, Powder Technology 46 (1986) 67–75.

[64] V. Mazel, V. Busignies, S. Duca, B. Leclerc, P. Tchoreloff, Original predictive

approach to the compressibility of pharmaceutical powder mixtures based on the

Kawakita equation, International Journal of Pharmaceutics 410 (2011) 92–98.

[65] V. Busignies, V. Mazel, H. Diarra, P. Tchoreloff, Prediction of the compressibility of

complex mixtures of pharmaceutical powders, International Journal of Pharmaceutics

436 (2012) 862–868.

[66] H. Leuenberger, R. Leu, Formation of a Tablet: A Site and Bond Percolation

Phenomenon, Journal of Pharmaceutical Sciences 81 (1992) 976–982.

[67] H.L. Martin Kuentz, A new theoretical approach to tablet strength of a binary mixture

consisting of a well and a poorly compactable substance, European Journal of

Pharmaceutics and Biopharmaceutics 49 (2000) 151–159.

[68] V. Busignies, B. Leclerc, P. Porion, P. Evesque, G. Couarraze, P. Tchoreloff,

Application of percolation model to the tensile strength and the reduced modulus of

elasticity of three compacted pharmaceutical excipients, European Journal of

Pharmaceutics and Biopharmaceutics 67 (2007) 507–514.

REFERENCES

170

[69] Martin Kuentz, Hans Leuenberger, M. Kolb, Fracture in disordered media and tensile

strength of microcrystalline cellulose tablets at low relative densities, International


[70] N. Ramirez, L.M. Melgoza, M. Kuentz, H. Sandoval, I. Caraballo, Comparison of

different mathematical models for the tensile strength-relative density profiles of

binary tablets, European journal of pharmaceutical sciences official journal of the

European Federation for Pharmaceutical Sciences 22 (2004) 19–23.

[71] M. Kuentz, H. Leuenberger, Modified Young's modulus of microcrystalline cellulose

tablets and the directed continuum percolation model, Pharmaceutical Development

and Technology 1 (1998) 13–19.

[72] K. Kawakita, K.-H. Lüdde, Some considerations on powder compression equations,

Powder Technology 4 (1971) 61–68.

[73] Fredrik Nicklasson, Göran Alderborn, Analysis of the Compression Mechanics of

Pharmaceutical Agglomerates of Different Porosity and Composition Using the Adams

and Kawakita Equations, Pharmaceutical Research (2000).

[74] J. Nordström, K. Welch, G. Frenning, G. Alderborn, On the physical interpretation of

the Kawakita and Adams parameters derived from confined compression of granular

solids, Powder Technology 182 (2008) 424–435.

[75] G. Frenning, J. Nordström, G. Alderborn, Effective Kawakita parameters for binary

mixtures, Powder Technology 189 (2009) 270–275.

[76] J.M. Sonnergaard, Impact of particle density and initial volume on mathematical

compression models, European Journal of Pharmaceutical Sciences (2000) 307–315.

[77] J. Sonnergaard, Investigation of a new mathematical model for compression of

pharmaceutical powders, European Journal of Pharmaceutical Sciences 14 (2001) 149–

157.

[78] J. Nordstrom, I. Klevan, G. Alderborn, A particle rearrangement index based on the

Kawakita powder compression equation, Journal of Pharmaceutical Sciences 98 (2009)

1053–1063.

[79] V. Busignies, V. Mazel, H. Diarra, P. Tchoreloff, Prediction of the compressibility of

complex mixtures of pharmaceutical powders, International Journal of Pharmaceutics

436 (2012) 862–868.

[80] C.M. Ramaswamy, Y. Varma, D. Venkateswarlu, Compaction of mixtures of

materials, The Chemical Engineering Journal 1 (1970) 168–171.

REFERENCES

171

[81] V. Mazel, V. Busignies, S. Duca, B. Leclerc, P. Tchoreloff, Original predictive

approach to the compressibility of pharmaceutical powder mixtures based on the

Kawakita equation, International Journal of Pharmaceutics 410 (2011) 92–98.

[82] C.-Y. Wu, S.M. Best, A.C. Bentham, B.C. Hancock, W. Bonfield, A simple predictive

model for the tensile strength of binary tablets, European Journal of Pharmaceutical

Sciences 25 (2005) 331–336.

[83] A. Michrafy, M. Michrafy, M.S. Kadiri, J.A. Dodds, Predictions of tensile strength of

binary tablets using linear and power law mixing rules, International Journal of

Pharmaceutics 333 (2007) 118–126.

[84] A. Gupta, G.E. Peck, R.W. Miller, K.R. Morris, Real-time near-infrared monitoring of

content uniformity, moisture content, compact density, tensile strength, and Young's

modulus of roller compacted powder blends, Journal of Pharmaceutical Sciences 94

(2005) 1589–1597.

[85] I. Klevan, Compression Analysis of Pharmaceutical Powders Assessment of

Mechanical Properties. PhD Thesis, University of Tromsø, 2011.

[86] S. Patel, A.M. Kaushal, A. Bansal, Kumar, Effect of particle size and compression

force on compaction behavior and derived mathematical parameters of compressibility,

Pharmaceutical Research 24 (2007) 111–124.

[87] J. Sonnergaard, Investigation of a new mathematical model for compression of

pharmaceutical powders, European Journal of Pharmaceutical Sciences 14 (2001) 149–

157.

[88] J.M. Sonnergaard, Quantification of the compactibility of pharmaceutical powders,


[89] J. Dressler, Vergleichende Untersuchungen pharmazeutischer Hilfsstoffe unter Einsatz

eines inkrementalen Weggebers zur präzisen Wegmessung an einer Exzenter-

Tablettenpresse. PhD thesis, University of Tübingen, 2002.

[90] H. Vromans, A.H. De Boer, G.K. Bolhuis, C.F. Lerk, K.D. Kussendrager and H.

Bosch, Studies on tableting properties of lactose, Pharmaceutisch Weekblad Scientific

Edition 7 (1985) 186–193.

[91] Coenraad F. Lerk, Consolidation and Compaction of Lactose, Drug Development and

Industrial Pharmacy 19 (1993) 2359–2398.

[92] A. Michrafy, M. Michrafy, M.S. Kadiri, J.A. Dodds, Predictions of tensile strength of

binary tablets using linear and power law mixing rules, International Journal of

Pharmaceutics 333 (2007) 118–126.

REFERENCES

172

[93] Busignies V., Leclerc B., Porion P., Evesque P., Couarraze G., Tchoreloff P.,

Compaction behaviour and new predictive approach to the compressibility of binary

mixtures of pharmaceutical excipients, European Journal of Pharmaceutics and


[94] J.M. Sonnergaard, Impact of particle density and initial volume on mathematical

compression models, European Journal of Pharmaceutical Sciences 11 (2000) 307–

315.

[95] M. Whiteman, Yarwood J. R., The Evaluation of Six Lactose-Based Materials as

Direct Compression Tablet Excipients, Drug Development and Industrial Pharmacy 8

(1988) 1023–1040.

[96] Busignies V., Leclerc B., Porion P., Evesque P., Couarraze G., Tchoreloff P.,

Investigation and modelling approach of the mechanical properties of compacts made

with binary mixtures of pharmaceutical excipients, European Journal of Pharmaceutics

and Biopharmaceutics 64 (2006) 51–65.

[97] N. Müller, Untersuchungen zur Prozessüberwachung und -regulierung bei der

Walzenkompaktierung mittels Drehmomenterfassung an der Granuliereinheit. PhD

thesis, University of Bonn, 2012.

[98] P.K. Haress Mangal, Experimental determination of residence time distribution in

continuous dry granulation, International Journal of Pharmaceutics (2017) 91–100.

[99] J. Sakwanichol, S. Puttipipatkhachorn, G. Ingenerf, P. Kleinebudde, Roll

compaction/dry granulation: comparison between roll mill and oscillating granulator in

dry granulation, Pharmaceutical Development and Technology 17 (2012) 30–39.

[100] J. Hilden, G. Earle, E. Lilly, Prediction of roller compacted ribbon solid fraction for

quality by design development, Powder Technology 213 (2011) 1–13.

[101] H. Vromans, F.C. Lerk, Densification properties and compactibility of mixtures of

pharmaceutical excipients with and without magnesium stearate 46 (1988) 183–192.

[102] G. Alderborn, K. Pasanen, C. Nystrom, Studies on direct compression of tablets. XL

Characterization of particle fragmentation during compaction by permeametry

measurements of tablets☆, International Journal of Pharmaceutics 23 (1985) 79–86.

[103] T. Jahn, K.-J. Steffens, Press chamber coating as external lubrication for high speed

rotary presses: lubricant spray rate optimization, Drug Development and Industrial

Pharmacy 31 (2005) 951–957.

REFERENCES

173

[104] I. Sinka, J. Cunningham, A. Zavaliangos, The effect of wall friction in the compaction

of pharmaceutical tablets with curved faces: A validation study of the Drucker–Prager

Cap model, Powder Technology 133 (2003) 33–43.

[105] F. Freitag, K. Reincke, J. Runge, W. Grellmann, P. Kleinebudde, How do roll

compaction/dry granulation affect the tableting behaviour of inorganic materials?

Microhardness of ribbons and mercury porosimetry measurements of tablets, European

journal of pharmaceutical sciences official journal of the European Federation for

Pharmaceutical Sciences 22 (2004) 325–333.

[106] N. Huber, Veränderung der Tablettiereigenschaften von Trockengranulaten aus

Mikrokristalliner Cellulose durch Zusatz verschiedener Hilfsstoffe. PhD thesis, 2011.

[107] L. Perez-Gandarillas, A. Mazor, D. Souriou, O. Lecoq, A. Michrafy, Compaction

behaviour of dry granulated binary mixtures, Powder Technology 285 (2015) 62–67.

[108] S. Hein, K.M. Picker-Freyer, J. Langridge, Simulation of roller compaction with

subsequent tableting and characterization of lactose and microcrystalline cellulose,


[109] K.A. Riepma, H. Vromans, K. Zuurman and CF. Lerk, The effect of dry granulation

on the consolidation and compaction of crystalline lactose, International Journal of

Pharmaceutics 97 (1993) 29–38.

[110] Sari Westermarck, Anne Mari Juppo, Lasse Kervinena, Jouko Yliruusib, Pore

structure and surface area of mannitol powder, granules and tablets determined with

mercury porosimetry and nitrogen adsorption, European Journal of Pharmaceutics and


[111] Juppo, A.M. and Yliruusi, J, Effect of Amount of Granulation Liquid on Total Volume

and Pore Size Distribution of Lactose, Glucose and Mannitol Granules, European

Journal of Pharmaceutics and Biopharmaceutics 40 (1994) 299–309.

[112] A.K. Samanta, K.Y. Ng, P.W.S. Heng, Cone milling of compacted flakes: Process

parameter selection by adopting the minimal fines approach, International Journal of

Pharmaceutics 422 (2012) 17–23.

[113] S. Yu, C.-Y. Wu, M.J. Adams, G. Reynolds, B. Gururajan, J. Gargiuli, T. Leadbeater,

R. Roberts, D.J. Parker, The use of positron emission particle tracking (PEPT) to study

milling of roll-compacted microcystalline cellulose ribbons, Powder Technology 285

(2015) 74–79.

[114] O. Simon, P. Guigon, Correlation between powder-packing properties and roll press

compact heterogeneity, Powder Technology 130 (2003) 257–264.

REFERENCES

174

[115] H. Busies, Dichteverteilung in Schülpen. PhD thesis, University of Bonn, 2006.

[116] A.R. Muliadi, J.D. Litster, C.R. Wassgren, Modeling the powder roll compaction

process: Comparison of 2-D finite element method and the rolling theory for granular

solids (Johanson's model), Powder Technology 221 (2012) 90–100.

[117] A.M. Miguélez-Morán, Roller compaction of pharmaceutical ingredients: On the

understanding of the compaction and the use of knowledge based Applications in the

formulation of tablets. PhD thesis, University of Heidelberg, 2009.

[118] A. Mazor, L. Perez-Gandarillas, A. de Ryck, A. Michrafy, Effect of roll compactor

sealing system designs: A finite element analysis, Powder Technology 289 (2016) 21–

30.

[119] J.D. Kirsch, J.K. Drennen, Nondestructive tablet hardness testing by near-infrared

spectroscopy: A new and robust spectral best-fit algorithm, Journal of Pharmaceutical

and Biomedical Analysis 19 (1999) 351–362.

[120] Marcelo Blanco, Manel Alcala, Josep M. Gonzalez, Ester Torras, A Process Analytical

Technology Approach Basedon Near Infrared Spectroscopy: Tablet Hardness,Content

Uniformity, and Dissolution Test Measurements of Intact Tablets 95 (2006) 2137–

2144.

[121] P.S. Jeckel, Bestimmung wesentlicher Tablettenparameter mit hilfe der Nahinfrarot-

spektroskopie. PhD thesis, University of Bonn, 2008.

[122] J. Austin, A. Gupta, R. Mcdonnell, G.V. Reklaitis, M.T. Harris, The use of near-

infrared and microwave resonance sensing to monitor a continuous roller compaction

process, Journal of Pharmaceutical Sciences 102 (2013) 1895–1904.

[123] A. Candolfi, D.L. Massarta, S. Heuerdingb, Investigation of sources of variance which

contribute toa proper measuring setup or data pre-process techniques NIR-

spectroscopic measurement of pharmaceutical formulations, Analytica Chimica Acta

345 (1997) 185–196.

[124] Asmund Rinnan, Frans van den Berg, Sören Balling Engelsen", Review of the most

common pre-processing techniques for near-infrared spectra.

[125] I. Habib, Near infra-red reflectance spectroscopic determination of metformin in

tablets, Talanta 60 (2003) 185–190.

[126] Xiangji Zhou, Patricia Hines, Matthew W. Borer, Moisture determination in

hygroscopic drug substances by near infrared spectroscopy, Journal of Pharmaceutical

and Biomedical Analysis 17 (1997) 219–225.

REFERENCES

175

[127] Soh, Josephine L. P., F. Wang, N. Boersen, R. Pinal, G.E. Peck, M.T. Carvajal, J.

Cheney, H. Valthorsson, J. Pazdan, Utility of multivariate analysis in modeling the

effects of raw material properties and operating parameters on granule and ribbon

properties prepared in roller compaction, Drug Development and Industrial Pharmacy

34 (2008) 1022–1035.

[128] Martin Kuentz, Hans Leuenberger, M. Kolb, Fracture in disordered media and tensile

strength ofmicrocrystalline cellulose tablets at low relative densities, International

Journal of Pharmaceutics (1999) 243–255.

[129] S. Patel, A.M. Kaushal, A.K. Bansal, Compression Physics in the Formulation

Development of Tablets, Critical Reviews in Therapeutic Drug Carrier Systems 23

(2006) 1–66.

[130] G. Thoorens, F. Krier, B. Leclercq, B. Carlin, B. Evrard, Microcrystalline cellulose, a

direct compression binder in a quality by design environment--a review, International


[131] A.P. Webb, An Introduction To The Physical Characterization of Materials by

Mercury Intrusion Porosimetry with Emphasis On Reduction And Presentation of

Experimental Data, 2001,

http://www.particletesting.com/Repository/Files/An_Introduction_to_the_Physical_Ch

aracterization_of_Materials_by_Mercury_Intrusion_Porosimetry.pdf, accessed 14

December 2016.

[132] E.W. Washburn, Note on a method of determining the distribution of pore sizes in a

porousmaterial, Proc.Nat.Acad.Sci. (1921) 115–116.

[133] Verena Maria Gläßer, Kompressionsanalyse im Vergleich zwischen hydraulischer und

Rundlauftablettenpresse. Diploma thesis, University of Tübingen, 2008.

[134] Gabriele Reich, Near-infrared spectroscopy and imaging: Basic principles and

pharmaceutical applications, Advanced Drug Delivery Reviews 57 (2005) 1109–1143.

[135] Celio Pasquini, Near Infrared Spectroscopy: Fundamentals, Practical Aspects and

Analytical Applications, Journal of the brazilian chemical society 14 (2003) 198–219.

[136] H.W. Siesler, Y.Ozaki, S. Kawata, H.M. Heise (Ed.), Near-infrared Spectroscopy -

Principles, Instruments, Applications, Wiley-Vch, 2002.

[137] Alexey L. Pomerantseva and Oxana Ye. Rodionova, Process analytical technology: a

critical view ofthe chemometricians, Journal of Chemometrics 26 (2012) 299–310.

[138] Sanni Matero, Frans van den Berg, Sami Poutiainen, Jukka Rantanen, Jari Pajander,

Towards Better Process Understanding: Chemometrics and Multivariate Measurements

REFERENCES

176

in Manufacturing of Solid Dosage Forms, Journal of Pharmaceutical Sciences 102

(2013) 1385–1403.

[139] Theodora Kourti, Application of latent variable methods to process control and

multivariate statistical process control in industry, International Journal of Adaptive

Control and Signal Processing 19 (2005) 213–246.

[140] A.K. Samanta, A.D. Karande, K.Y. Ng, Heng, Paul Wan Sia, Application of near-

infrared spectroscopy in real-time monitoring of product attributes of ribbed roller

compacted flakes, AAPS PharmSciTech 14 (2013) 86–100.

[141] Waltraud Kessler, Multivariate Datenanalyse für Pharma-, Bio- und Prozessanalytik, 1

ed., Wiley-Vch, 2007.

[142] S.M. Short, R.P. Cogdill, P.L.D. Wildfong, J.K.3. Drennen, C.A. Anderson, A near-

infrared spectroscopic investigation of relative density and crushing strength in four-

component compacts, Journal of Pharmaceutical Sciences 98 (2009) 1095–1109.

Documents

DRY GRANULATION VIA ROLLER COMPACTION : INVESTIGATION …hss.ulb.uni-bonn.de/2018/5074/5074.pdf · However, scaling up roller compaction processes is still not fully understood, identifying